🧭 Iterion feature inventory
140 criteria · 16 categories · 35 bots in the repository
September 11, 2026 · Capabilities, use cases and limitations of the reviewed version.
Iterion combines agent orchestration and Git workflows with active supervision, memory across missions, asynchronous interaction, file recovery, plugins, scheduling and team coordination. Use this inventory to identify the capabilities your developers and operators need, and check their scope before adoption.
← Comparison guide · Feature availability — 10 products
Status key. ✅ Documented in the reviewed version. 🟡 Available with a material restriction on backend, deployment, scope or activation. 🧩 A method supplied as a bot using engine capabilities. Deferred work is listed separately and excluded from the total.
Evidence. Documentation review and targeted code inspection: 139 criteria at commit
bbc1ddc7844b81582b6dfc4d0b7f381680bb284b, plus connector criterion X09 at603d2a1e3314252dd2995e3bfaa4a7394e13093b; no execution tests were performed for this inventory. A checkmark does not imply parity across backends, managed availability or certified reliability. The project describes itself as experimental. current-state
⭐ Capabilities to explore
| Capability | What it enables in your workflow |
|---|---|
| 👁️ Active supervision | An agent can monitor a mission and send guidance while work is in progress. |
| 🙋 Asynchronous questions | An agent can keep working while you prepare an answer, then wait at a chosen point. |
| 📚 Memory and knowledge | Selected knowledge can be reused across runs, bots and projects. |
| ⏪ Recovery and rewind | Resume the graph, investigate errors and restore supported file states. |
| 🧩 Skills and plugins | Distribute team methods, tools and technical knowledge. |
| ⚡ Triggers and scheduling | Start a mission from a forge event, a schedule, a card or another result. |
| 📋 Board and shared configuration | Track work and delegate specific configuration fields to business users. |
| 🏢 Native multi-tenant platform | Operate agent workflows across organizations and teams with scoped resources, credentials and roles. |
| 🔐 Permissions and data protection | Make access, secrets and data transformations explicit workflow settings. |
These categories help you assess fit. Their presence in Iterion does not establish their absence elsewhere.
🗺️ Browse the inventory
📝 Authoring and sharing · 🔀 Orchestration · 🧠 Models, sessions and context · 🙋 Human interaction · 👁️ Supervision · 📚 Memory and knowledge · 🧩 Skills, plugins and integrations · ⏪ Recovery and rewind · 🌿 Git and delivery · ⚡ Triggers and scheduling · 📋 Team coordination · 🔐 Permissions, secrets and data · 🏢 Organizations, access and usage · 🔎 Observability · ⚙️ Deployment and evaluation · 🤖 Bot methods
📝 Authoring and sharing workflows
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| A01 | Declarative .bot workflows | ✅ | Define agents, tools, data and transitions in a versionable file. A deterministic graph does not make LLM responses identical. dsl |
| A02 | Synchronized visual and source editors | ✅ | Canvas, node library, inspector, live validation, undo and redo. visual-editor |
| A03 | Guided bot creation | ✅ | Starter templates, mission form, parameters and a first run; equivalent creation through the CLI. visual-editor |
| A04 | Portable .botz bundles | ✅ | Deterministic ZIP archive containing the workflow, prompts, skills, attachments and manifest; supports declaring a minimum engine version. bundles |
| A05 | Reusable variables, presets and prompts | ✅ | Parameterize a method at launch and distribute specializations without copying the entire workflow. bundles dsl |
| A06 | Team bot editing | ✅ | Create, duplicate and edit bundle files in cloud Studio, with conflict checks on save. visual-editor |
| A07 | Importing an existing Claude workflow | 🟡 | Static converter from Claude JavaScript files to a .bot draft, with a report of unsupported elements. Does not import n8n workflows. import |
| A08 | Schemas and static diagnostics | ✅ | Validate references, properties and graph/data-flow constraints before execution. General node-output schema validation is an engine option, not enabled by product entry points in this build; compute outputs are checked separately. dsl output-validation-code compute-types-code |
| A09 | Mermaid diagram export | ✅ | Generate a compact, detailed or full workflow diagram for documentation and review. cli-reference |
🔀 Orchestrating agents, code and events
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| O01 | Explicit agents and judges | ✅ | Separate production from evaluation in the graph; a judge is a configurable LLM evaluation, not proof of correctness. dsl |
| O02 | Deterministic tools and scripts | ✅ | Command nodes or JavaScript, Python and shell/Bash scripts, with stdout capture and error handling. dsl |
| O03 | Computation without a model or shell | ✅ | The compute node provides bounded expressions, transformations, map/filter/reduce and typed outputs. dsl |
| O04 | Conditional, LLM and round-robin routing | ✅ | Select a branch using rules, a model decision or a declared rotation. routers |
| O05 | Parallel branches and result collection | 🟡 | fan_out_all, fan_out_each, with wait_all or best_effort collection; concurrent workspace writes are restricted. routers groups-iteration-subbots |
| O06 | Bounded, data-driven loops and ordered iteration | ✅ | Fixed or data-resolved bounds, foreach, and explicit unbounded loops governed by fuel and a stagnation monitor. Empty foreach input requires a guard if its body must never run. groups-iteration-subbots loop-fuel-code loop-edges-code |
| O07 | Reusable parameterized groups | ✅ | Declare a subgraph and instantiate it with distinct parameters and prefixes. groups-iteration-subbots |
| O08 | Sub-bots with their own runs | 🟡 | Call another bot, retrieve its output and isolate its work; budgets, memory and supervisors have parent/child limitations. groups-iteration-subbots memory-and-knowledge supervisors |
| O09 | Resource pools and concurrency limits | ✅ | Declare resources and reserve a slot with needs, for example to limit simultaneous worktrees. groups-iteration-subbots dsl |
| O10 | Emitting and waiting for internal events | 🟡 | In-run emit/wait, retained events for late waiters and mandatory timeouts. Durable parking on external events is deferred. event-primitives |
🧠 Models, sessions and context
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| B01 | Six execution backends | 🟡 | claw, claude_code, pi, codex, kimi, grok, with different capabilities and support levels. backends delegation |
| B02 | Model and backend selection per node | ✅ | Mix executors in one workflow, select the model independently of the backend and apply launch overrides. backends |
| B03 | Provider or credential fallback | 🟡 | Configurable fallback chains when the initial access is unsuitable; changes should remain observable. backends |
| B04 | Cross-backend fallback and skip routes | 🟡 | Declare a fallback executor or a workflow-defined skip route; prerequisites are validated. Codex is excluded from the first cross-backend fallback version. backends |
| B05 | Fresh, inherited or forked sessions | 🟡 | Choose conversational continuity between steps; session resume and fork are not wired for Kimi/Grok. dsl delegation |
| B06 | Persistent sessions on node re-entry | 🟡 | session: persist reuses a node's conversation in a loop; supported on Claude Code, pi and Codex along the main workflow path. session-persist |
| B07 | Artifact-only handoff | ✅ | artifacts_only passes the useful result without carrying the previous step's conversation history. dsl |
| B08 | Context compaction | 🟡 | Configurable threshold and retention of recent exchanges; depends on the executor and can support error recovery. dsl run-recovery |
| B09 | Behavior dials | ✅ | Qualitative or numeric prompt settings for depth, caution, style and more. They guide the model without enforcing permissions or guaranteeing quality. cursors |
| B10 | Reasoning effort settings | 🟡 | Choose reasoning_effort per node or through input data; levels adapt to model and backend support, independently of behavior dials. ultracode effort-code |
| B11 | Sub-agents and workflows created within a node | 🟡 | ultracode enables dynamic orchestration on claw and Claude Code depending on model, harness and tools; separate from declared fan-out and sub-bots. Pi does not provide this capability. ultracode subagents-code pi-code |
🙋 Human interaction and guidance
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| H01 | Typed approval forms | ✅ | human nodes support text, choices, booleans, numbers and lists; review context is separate from expected answers. human-in-the-loop |
| H02 | Agent questions during execution | 🟡 | ask_user with answers through Studio, CLI or API; support depends on backend and transport. human-in-the-loop delegation |
| H03 | Questions without blocking the agent | 🟡 | ask_user_async lets the agent continue before receiving answers; asynchronous interaction runs on the main path, outside fan-out branches. async-interaction |
| H04 | Explicit answer synchronization | 🟡 | await_answers synchronizes pending questions; the node requires a timeout, and the run remains active while waiting. async-interaction |
| H05 | Automatic answers or human escalation | 🟡 | llm and llm_or_human modes follow workflow policy; an automated answer is not human approval. human-in-the-loop |
| H06 | Guidance added to an active run | 🟡 | Inbox and steering redirect work at backend-defined intake points; they do not instantly cancel an ongoing action. backends supervisors |
| H07 | Files and images at launch or in an answer | 🟡 | Persisted attachments with MIME type, size and hash; vision depends on the model. Review gates do not accept the same uploads as ordinary human nodes. attachments human-in-the-loop |
| H08 | Review dialogue before integration | 🟡 | Review companion, correction requests and an optional review environment; explicitly choose human_required or an agent verdict. review-merge-gate |
| H09 | Operator pause with checkpoint | 🟡 | Request a pause at the next safe boundary, then resume; local in-process path only, with remote cloud pause unimplemented. pause-code |
👁️ Supervising an active mission
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| S01 | Agent supervising another agent | 🟡 | Observe an agent/judge node and inject guidance during execution; supervisors declared inside sub-bots are not currently wired. supervisors |
| S02 | Supervision monitors, cadence and budget | ✅ | React to events, errors or cost signals with a cooldown and evaluation limit; account for supervision costs. supervisors |
| S03 | Attaching a supervisor to a session | 🟡 | Attach through the CLI to a local run, or to a Claude Code session with prepared transcript and hooks; remote cloud attachment is deferred. supervisors |
| S04 | Changing budgets and iterations during a run | 🟡 | Raise limits or grant extra iterations with acknowledgment and a recorded event; changes apply at safe boundaries. dsl steering-code |
📚 Reusable memory and knowledge
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| M01 | Document memory across runs | ✅ | Retain knowledge as Markdown documents beyond a single run's artifacts. memory-and-knowledge |
| M02 | Seven visibility scopes | 🟡 | Private run, bot, project, multiple projects, user, organization and global catalog; storage and context determine access. Global scope is read-only for organizations. memory-and-knowledge |
| M03 | Agent memory read, write and list tools | 🟡 | memory_read, memory_write, memory_list, with configurable permissions; check the adapter and backend in use. memory-and-knowledge memory-code |
| M04 | Index and selective context loading | ✅ | Inject an index, preload selected documents and reinject memory before compaction according to configuration. memory-and-knowledge |
| M05 | Automatic bot memory | 🟡 | auto_memory maintains a MEMORY.md per bot and repository; opt-in on claw/Claude Code/pi, with restrictions for cloud sandboxes without read-back. memory-and-knowledge |
| M06 | Memory quotas and secret detection | 🟡 | Aggregate, space and document quotas; structured secret detection does not cover all confidential data. memory-and-knowledge |
| M07 | Memory import and export | ✅ | tar.gz archives through CLI/API, with skip, overwrite or rename strategies to retain or transfer knowledge. memory-and-knowledge |
🧩 Skills, plugins and technical integrations
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| X01 | Editable skills library | 🟡 | Global, project or bundled skills; management CLI and local editor, with selected skills transported to cloud runners. skills-library |
| X02 | Skills selected per workflow or node | ✅ | Name and description enter the context; content loads on demand. A missing skill produces a warning. skills-library |
| X03 | Plugins with declarative contributions | 🟡 | MCP, skills, output rewriting, commands, agents, hooks and lifecycle features; backend support varies by contribution. plugins |
| X04 | Private organization plugins from Git | ✅ | Resolve private and public sources, then transport them for local or cloud runs. plugins |
| X05 | Command output compression | 🟡 | RTK plugin and compress modes reduce text sent to the model; plugin availability and activation depend on the interface in use. plugins |
| X06 | MCP client in workflows | 🟡 | Declare MCP servers, inherit them or disable them per node; transports and support depend on the backend. dsl delegation |
| X07 | MCP server for controlling Iterion | ✅ | iterion mcp exposes launch, monitoring, human answers and board operations, locally or against a configured remote instance. mcp-server |
| X08 | Web search and page reading | 🟡 | Native tools or configured providers/MCP, including DuckDuckGo, Brave, SearXNG and Firecrawl; coverage varies by backend. web-search |
| X09 | Deterministic connector actions and package generation | 🟡 | Local in-process action: nodes use typed parameters, bound connections and structured results with pagination/completion indicators. Generate packages from OpenAPI/Swagger with an authored overlay. Project catalogs require opt-in; cloud transport and the catalog MCP facade are outside this delivered scope. connector-dsl connector-local-code connector-cli-code |
⏪ Recovery, repair and rewind
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| R01 | Workflow state checkpoints | 🟡 | Resume eligible states with outputs, counters and position; interrupted steps may execute again. resume |
| R02 | Sub-bot recovery after restart | 🟡 | Reattach the parent to its recorded child run instead of always launching a new child, when the state is recoverable. subbot-reattach |
| R03 | Rewinding to an earlier node | 🟡 | rewind invalidates dependent outputs; --auto locates a changed node. Branch and starting-state restrictions apply. resume |
| R04 | File snapshots outside Git history | 🟡 | Filesystem tracker wired to local CLI/Studio, with deduplicated states at execution boundaries; cloud runners do not wire this tracker. Separate from pushed Git workspace checkpoints. workspace-versioning tracker-wiring-code runner-code |
| R05 | Targeted restoration and pre-rewind backup | 🟡 | Restore run-produced files from available local snapshots while protecting operator changes; ignored or oversized paths are excluded. workspace-versioning resume |
| R06 | Actions with verifiable postconditions | ✅ | Check a condition before and after a command, skip an already-satisfied action and select required/recover/best_effort policy. verified-actions verified-code |
| R07 | Bounded action repair | 🟡 | Propose a corrected command, then optionally delegate recovery to an agent; opt-in, bounded attempts, with success determined by the postcondition. verified-actions verified-code |
| R08 | Controlled retries and automatic recovery | 🟡 | Error classification, backoff, compaction or human pause; optional CLI auto-resume excludes terminal errors. run-recovery recovery-code |
| R09 | Recovery hint after pod loss | 🟡 | Expose the last successfully pushed workspace checkpoint and a recovery command; this does not automatically restore files or validate a deliverable. resume |
| R10 | Forking a run from an earlier exchange | 🟡 | Create a resumable run with modified inputs and restore the code snapshot when available; conversational precision depends on the backend. cli-reference fork-code |
🌿 Working and delivering in a repository
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| G01 | Separate worktree per run | 🟡 | Isolate a run's Git changes; behavior depends on mode and repository, particularly for empty repositories. dsl repo-scope worktree-finalization |
| G02 | Git finalization policy | 🟡 | Retain or integrate the result according to the selected policy; finalization authority must be granted. cli-reference worktree-finalization |
| G03 | Merge gate with result review | 🟡 | Review, corrections and merge/squash decision within the supported flow; configure acceptance criteria and the target branch. review-merge-gate |
| G04 | Changes and commits view | ✅ | Inspect modified files, diffs and commits from a run; metadata is persisted for documented cloud paths. visual-editor cloud-git |
| G05 | Conservative worktree pool cleanup | 🟡 | Bounded collection of eligible old worktrees preserves dirty, active or recoverable work; it is not a strict disk cap. worktree-pool |
| G06 | Devbox tooling environment | 🟡 | Prepare tooling declared by the bot and repository; requires a compatible installation and execution environment. devbox |
| G07 | Repository targeting, connection and creation | 🟡 | Guided forge connection, target selection and authorized creation in cloud Studio. Manifests also support optional or no repository requirements; creation needs forge permissions. repo-scope |
⚡ Triggers, scheduling and chaining
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| T01 | Manifest-defined invocations | ✅ | Declare forge, command, schedule, board and keepalive inputs, with direct launch or card creation. bot-invocations |
| T02 | Multi-forge and JSON inbound webhooks | ✅ | GitHub, GitLab, Forgejo/Gitea or generic payloads; configurable event/project/author filters and authentication. webhooks |
| T03 | Forge commands and conversations | 🟡 | Slash commands and bot replies in authorized GitHub/GitLab threads; replies still rely on a skill, while native forge.reply is deferred. forge-conversations bot-invocations |
| T04 | Local and cloud scheduling | ✅ | Cron, guards and overlap policy; runs through host cron locally or the platform scheduler. scheduling |
| T05 | Keepalive and availability waits | 🟡 | Relaunch recurring missions using guards and availability windows; configure behavior per bot and backend. scheduling bot-invocations |
| T06 | Triggers on events and run outcomes | ✅ | Subscribe to board events, explicit events or run completion to chain missions. trigger-spine |
| T07 | Durable outbox for cloud board effects | 🟡 | Materialization, claim, retry and permanent failure for trigger effects; at-least-once execution with documented residual failure windows. trigger-outbox |
| T08 | Automatic mission outcome routing | 🟡 | Policy frozen at launch for merge, relaunch or escalation; server option and decision log. outcome-router |
| T09 | HTTP callbacks and web notifications | 🟡 | HMAC-signed completion callbacks, operator alerts and Web Push require configuration; native desktop notifications are deferred. outbound-callbacks notifications usage-caps |
📋 Coordinating team work
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| K01 | Native kanban independent of a forge | ✅ | Cards, columns, labels, priorities, assignments and comments through Studio or CLI. native-tracker |
| K02 | Typed custom fields and dependencies | ✅ | Validated card schema, hard blockers and waiting_deps state prevent premature launches. native-tracker |
| K03 | Autonomous issue dispatcher | 🟡 | Selection, claim, launch, retries, stall detection and concurrency limits; native tracker and configurable adapters. dispatcher |
| K04 | Pipeline board and mission control | ✅ | View planned tasks and active runs, launch by priority and limit concurrent pipelines. visual-editor native-tracker |
| K05 | GitHub Projects V2 synchronization | 🟡 | Bidirectional statuses; issue fields synchronize according to a specific mapping, not arbitrary field replication. github-board-sync |
| K06 | Links between cards, runs and results | ✅ | Latest run, worktree, wait states and event history establish work provenance. native-tracker |
| K07 | Configuration editor for non-operators | 🟡 | Share revocable access to selected repository file fields; forge commits, conflict checks and audit, without operator Studio access. config-share |
🔐 Permissions, secrets and data
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| P01 | Tool permissions and read-only mode | 🟡 | Ask/deny gates on agent calls: claw, Claude Code and pi RPC; Kimi/Grok deny only, Codex without this gate. permission: does not apply to shell tool nodes. permissions delegation |
| P02 | Execution sandbox | 🟡 | Configurable isolation by backend, driver and deployment; auto can degrade, while Codex uses its own sandbox. sandbox |
| P03 | Bot-bound secrets and host controls | 🟡 | Bindings limit supplied secrets; allowed_hosts covers supported egress paths. This guarantee does not extend to every shell call. secrets-reference |
| P04 | Secrets sealed at rest | ✅ | AES-256-GCM with record binding; authorized cloud components share the key. Automatic master-key rotation is not currently available. secrets-reference |
| P05 | File-based secret injection | 🟡 | Materialize a secret with as: file, with in-run refresh on supported paths. secrets-reference file-secret-refresh |
| P06 | Local sensitive-data detection and masking | 🟡 | Go-based privacy_filter covers accounts, emails, phone numbers, URLs and secrets. Person names, postal addresses and dates are excluded. privacy_filter |
| P07 | Restoring masked values | 🟡 | privacy_unfilter restores original values if the vault file is retained; the vault is not replicated to Mongo/S3. Trace masking targets privacy_filter inputs and privacy_unfilter outputs, not data flowing elsewhere. privacy_filter vault-code runner-code |
| P08 | Upload and browser preview controls | 🟡 | MIME, size, path and SSRF controls on covered routes; active content is restricted, without a general guarantee that files are harmless. attachments browser-pane |
| P09 | System credential vault for desktop | 🟡 | macOS Keychain, Windows Credential Manager or Linux Secret Service integration requires the system service. Separate from cloud secret sealing. desktop keychain-code |
🏢 Organizations, access and usage
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| E01 | Multi-tenant organizations, teams and roles | ✅ | Separate platform resources by tenant and assign operator/administrator permissions and editing capabilities. cloud-overview config-share |
| E02 | Organization SSO | ✅ | Generic OIDC and dedicated connectors, with organization-specific configuration and access policies. sso |
| E03 | User or organization LLM keys | ✅ | BYOK, defaults and launch/integration overrides follow the resolution chain; providers bill separately. byok secrets-reference |
| E04 | Subscription/OAuth authentication | 🟡 | Reuse authentication supported by CLI backends within the limits of those accounts and their providers. oauth-forfait backends |
| E05 | Opt-in contributor credential pool | 🟡 | Credential lending with audience, availability, limits and per-run leases; does not increase provider quotas. credential-pool |
| E06 | Run budgets: cost, tokens, duration and iterations | 🟡 | Configurable limits, backend-dependent estimates, finalization headroom and separate sub-bot budgets; not a universal billing cap. dsl |
| E07 | Platform quotas and launch admission | 🟡 | Concurrency, rate, monthly and cost quotas; quota reads can allow a launch through on storage errors. quotas-and-limits |
| E08 | Subscription usage-window protection | 🟡 | Adjustable soft/hard thresholds and policies; measurement is currently specific to Claude Code and separate from monetary budgets. usage-caps |
| E09 | Personal tokens and administration audit | ✅ | Attributed API access and administrative actions; change history covers supported operations. cloud-rest-api secrets-reference |
| E10 | Effective settings and their origins | ✅ | Launch displays backend, permission, compression and auto-memory values and provenance, plus nodes with their own settings. settings-precedence |
🔎 Observing and inspecting results
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| U01 | Live run console | ✅ | Graph, events, logs and counters for monitoring execution in Studio. visual-editor |
| U02 | Artifacts and history per step/iteration | ✅ | Versioned outputs and persisted events for inspecting completed runs. persisted-formats |
| U03 | Cost reports and usage statistics | ✅ | Breakdown by model/provider, run counts, failures and P50/P95 durations; precision depends on available accounting. visual-editor |
| U04 | Session task list | 🟡 | Deterministic TodoWrite/todo_write history reconstructed from events; currently populated by Claude Code and claw. session-board |
| U05 | AI-generated monitoring widgets | 🟡 | Notes, metrics, checklists, progress and charts; local opt-in with an evaluation budget, not currently wired to cloud. session-board |
| U06 | Web previews and timestamped captures | 🟡 | Result URLs, run-linked captures and live browser through manual local attachment; cloud live browsing and Playwright auto-attach are deferred. browser-pane |
| U07 | Browsable file deliverables | ✅ | A tool can publish a generated file as a run attachment using a stdout directive; publication failures are reported without failing the tool. attachments |
| U08 | Post-mortem worktree terminal | 🟡 | Inspect an idle run in its retained environment; local Unix only, without a cloud host terminal. post-mortem-shell |
| U09 | Structured logs, errors and tracing | 🟡 | Process JSON, alerts and Sentry/GlitchTip integration; a DSN is required and tracing is enabled separately. observability |
| U10 | Prometheus, OTLP and Grafana dashboards | 🟡 | Metrics endpoint and supplied observability stack; costs, tokens, retries and durations depend on data exposed by each backend. observability-stack |
| U11 | Direct run file editing | 🟡 | Edit a text file in a retained local worktree; 4 MiB limit, without Git staging, locking or concurrent conflict checks. Separate from the bot editor. run-file-editor |
⚙️ Deploying and evaluating the platform
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| D01 | CLI, Studio, desktop app and self-hosted cloud | ✅ | Multiple interfaces around the same engine; feature parity is incomplete. current-state desktop |
| D02 | Distributed runners and platform storage | ✅ | NATS queue, runners, MongoDB and S3 storage; Kubernetes architecture and scaling require operation. cloud-architecture |
| D03 | Bot sources frozen at cloud launch | ✅ | Hashed snapshot of parent, sub-bots and resources; runners do not silently substitute their own catalog version. bot-bundle-snapshots |
| D04 | API and TypeScript SDK | ✅ | Control runs, resume and consume events; documented SDK support for Node, Deno and Bun. cloud-rest-api readme |
| D05 | Recipes and variant comparison | ✅ | Combine workflow, parameters, prompts, budgets and evaluation policy into benchmark campaigns. recipes |
| D06 | Convergence analysis on recorded runs | 🟡 | bench asymptote aggregates verdicts by iteration without replaying LLMs; it neither guarantees convergence nor constitutes a completed competitive benchmark. asymptote-bench |
| D07 | MIT-licensed source code | ✅ | Modifiable, self-hostable engine; third-party components and services retain their own terms. license |
| D08 | Desktop multi-project switching | ✅ | Select and return to multiple projects in the native app; onboarding prepares a first project. desktop desktop-projects-code |
| D09 | Desktop updates with signature verification | 🟡 | Verify the manifest and artifact using Ed25519, then offer installation according to package and platform. This does not imply macOS notarization or Windows signing. desktop updater-code |
🤖 Methods supplied as bots
| ID | Feature | Status | Usage and scope |
|---|---|---|---|
| C01 | Application and feature development | 🧩 | app-dev, feature-dev, feature-gap-fill, bmady bundles; workflows adaptable to the repository and acceptance criteria. bot-app-dev bot-feature-dev bot-feature-gap-fill bot-bmady |
| C02 | Code review and review dialogue | 🧩 | review-pr, revi-converse and review helpers; verify forge coverage and permissions. bot-review-pr bot-revi-converse |
| C03 | Continuous improvement and modernization | 🧩 | whole-improve-loop, branch-improve-loop, modernize, evolve; each bot defines its checks and approval steps. bot-whole-improve-loop bot-branch-improve-loop bot-modernize bot-evolve |
| C04 | Tests, coverage and instrumentation | 🧩 | test-coverage, e2e-coverage, golden-master, instrument build or inspect evidence suited to the project. bot-test-coverage bot-e2e-coverage bot-golden-master bot-instrument |
| C05 | Security audits and dependency maintenance | 🧩 | sec-audit-source, sec-audit-deps, dep-update-guard, secured-renovacy, supply-shield and variants; an automated audit cannot certify the absence of vulnerabilities. bot-sec-audit-source bot-sec-audit-deps bot-dep-update-guard bot-supply-shield |
| C06 | Vulnerability monitoring | 🧩 | vuln-watch detects issues and routes remediation according to configuration. bot-vuln-watch |
| C07 | Accessibility | 🧩 | rgaa-audit, ultra11y; assess audits and fixes against the tested scope, without assuming automatic compliance. bot-rgaa-audit bot-ultra11y |
| C08 | Documentation and architecture | 🧩 | docs-refresh, product-docs, wiki-gen, adr-cartograph, adr-rechallenge produce and update documents for review. bot-docs-refresh bot-product-docs bot-wiki-gen bot-adr-cartograph bot-adr-rechallenge |
| C09 | Content monitoring, triage and planning | 🧩 | feed-watch, issue-triage, whats-next, campaign; collection, synthesis and coordination using the supplied workflows. bot-feed-watch bot-issue-triage bot-whats-next bot-campaign |
🚧 Deferred work and remaining limitations
| Area | Status in this inventory |
|---|---|
dsl: 2 versioning, explicit file imports and migration | Described in ADR-098; no implementation evidence found in the parser/CLI at the reviewed commit. These commands and imports are not listed as shipped. dsl-roadmap |
| Durable run parking on external events | Deferred; distinct from triggers that launch runs and active internal waits. event-primitives async-interaction |
| Supervisors declared inside sub-bots | Declarations are not wired; parent support does not extend to children. supervisors |
| Cloud supervisor attachment and Studio toggle | Deferred; workflow-declared supervision exists, while CLI attachment targets local execution. supervisors |
| Memory parity across execution paths | The UI incorrectly lists some bot/project spaces; cloud sub-bots do not receive the parent's MemoryStore; driver read-back limits auto-memory. memory-and-knowledge |
| Filesystem snapshots and PII vault on another pod | The file tracker is local; the PII vault is a file in the executor's storeDir. Remote Git checkpoints and Mongo/S3 storage do not transfer them. tracker-wiring-code vault-code runner-code |
| Cloud live browser and Playwright auto-attach | Documented as future work. Existing previews and captures do not establish remote live browsing. browser-pane |
| AI-generated session widgets in cloud | Dedicated cloud storage and DSL declaration are deferred; current curation is local and opt-in. session-board |
| PII detection for names, postal addresses and dates | Planned for a later version; the five current categories provide limited coverage. privacy_filter |
| General node-output validation in product launches | The engine exposes opt-in validation and bounded correction, but CLI/Studio/cloud entry points do not enable it in this build. compute output conformance is active independently. output-validation-code output-options-code compute-types-code |
| Automatic master-key rotation and per-tenant KMS | Unavailable; current sealed storage does not provide this capability. secrets-reference |
| Native desktop notifications | Deferred; Web Push is separate from native application notifications. notifications |
| Automatic remediation bot triggered by Sentry | Described as an intention; error-watch does not yet exist. Error reporting and Sentry MCP access are separate capabilities. sentry-feedback-loop |
| Broader business connector catalog | The local deterministic foundation is available in X09. Broader packaged coverage, the catalog MCP facade and cloud transport remain outside this delivered scope; future connectors are not counted as installable. Tracking: connector catalog, issue #1072. |
Cross-cutting limitations. Sub-bot budgets are not aggregated into the parent's budget, and launch quotas are not a strict financial barrier during storage errors. Resuming a run does not guarantee exactly-once execution of external effects. dsl quotas-and-limits resume
Improvement methods. Bots can incrementally retain validated work and measure iteration verdicts. Outcomes depend on the configured checks; improvement on every iteration and freedom from regressions are not guaranteed. improvement-ratchet
🔬 Inventory evidence
Stable criterion IDs support future comparisons. The total spans primitives, interfaces, operations and methods; it is not a competitive score.
Row references point to a pinned repository version. Targeted code inspection complemented the documentation for the following mechanisms:
| Category | Evidence in the inspected code |
|---|---|
| Local connector actions | Typed action recipe and result contract, in-process resolver wiring, OpenAPI/Swagger generation: DSL, local launch wiring, CLI. Reviewed at the separate connector commit recorded in X09. |
| Loop control | Fixed/data-resolved bounds, finite fuel and loop stagnation checks: bounds and monitor, back-edge handling. |
| Output schemas | Opt-in general validation versus active compute conformance: option, validator, compute path. |
| Asynchronous interaction | StoreAsyncAskBinder, InteractionKindAsync persistence, human_input_requested event: async code. |
| Supervision | Coordinator, cooldown, MaxEvals limit and injection: supervision code. |
| Memory | Seven Visibility values, MemoryStore contract and tool/index wiring: scopes, contract, tools. |
| Verified actions | Skip precondition, recipe, postcondition and bounded repair stages: implementation. |
| Recovery | Error-class dispatch, attempt counter, retry/compaction/pause: implementation. |
| Files and rewind | Capture, Restore, RestoreOnly and protected paths: tracker contract. |
| Plugins | Manifest and typed contributions: implementation. |
| Shared editor | Readable field projection, patch validation and SHA conflict rejection: service. |
| Credential pool | Acquisition, contributor ranking, leases and release: broker. |
| Board triggers | Claim, retry and effect materialization: worker. |
| Limit adjustments | bump_loop/raise_budget, deduplication and typed replies: transport. |
| PII masking | Registration of both tools, validation and per-run vault: implementation. |
Documentation and targeted implementation checks underpin this inventory. Capabilities remain tied to the cited commit and should be reassessed when the version changes.
📦 The 35 source bundles
Counted from bots/*/main.bot at the reviewed commit. These are source bundles; not every binary embeds all 35, and some support other workflows.
adr-cartograph · adr-rechallenge · app-dev · arbitrate · bmady · branch-improve-loop · campaign · dep-update-guard · devbox-setup · docs-refresh · e2e-coverage · evolve · feature-dev · feature-gap-fill · feed-watch · golden-master · instrument · issue-triage · modernize · product-docs · revi-converse · review-env · review-pr · rgaa-audit · sec-audit-deps · sec-audit-source · secured-renovacy · supply-shield · supply-shield-cve · test-coverage · ultra11y · vuln-watch · whats-next · whole-improve-loop · wiki-gen
The CSV inventory contains the same criteria, statuses, limitations and references for further comparisons.
📎 Pinned sources
The short references above resolve to the following documents and files.
- asymptote-bench :
docs/asymptote-bench.md. - async-code :
pkg/backend/model/async_ask.go. - async-interaction :
docs/async-interaction.md. - attachments :
docs/attachments.md. - backends :
docs/backends.md. - bot-adr-cartograph :
bots/adr-cartograph/main.bot. - bot-adr-rechallenge :
bots/adr-rechallenge/main.bot. - bot-app-dev :
bots/app-dev/main.bot. - bot-bmady :
bots/bmady/main.bot. - bot-branch-improve-loop :
bots/branch-improve-loop/main.bot. - bot-bundle-snapshots :
docs/bot-bundle-snapshots.md. - bot-campaign :
bots/campaign/main.bot. - bot-dep-update-guard :
bots/dep-update-guard/main.bot. - bot-docs-refresh :
bots/docs-refresh/main.bot. - bot-e2e-coverage :
bots/e2e-coverage/main.bot. - bot-evolve :
bots/evolve/main.bot. - bot-feature-dev :
bots/feature-dev/main.bot. - bot-feature-gap-fill :
bots/feature-gap-fill/main.bot. - bot-feed-watch :
bots/feed-watch/main.bot. - bot-golden-master :
bots/golden-master/main.bot. - bot-instrument :
bots/instrument/main.bot. - bot-invocations :
docs/bot-invocations.md. - bot-issue-triage :
bots/issue-triage/main.bot. - bot-modernize :
bots/modernize/main.bot. - bot-product-docs :
bots/product-docs/main.bot. - bot-revi-converse :
bots/revi-converse/main.bot. - bot-review-pr :
bots/review-pr/main.bot. - bot-rgaa-audit :
bots/rgaa-audit/main.bot. - bot-sec-audit-deps :
bots/sec-audit-deps/main.bot. - bot-sec-audit-source :
bots/sec-audit-source/main.bot. - bot-supply-shield :
bots/supply-shield/main.bot. - bot-test-coverage :
bots/test-coverage/main.bot. - bot-ultra11y :
bots/ultra11y/main.bot. - bot-vuln-watch :
bots/vuln-watch/main.bot. - bot-whats-next :
bots/whats-next/main.bot. - bot-whole-improve-loop :
bots/whole-improve-loop/main.bot. - bot-wiki-gen :
bots/wiki-gen/main.bot. - browser-pane :
docs/browser-pane.md. - bundles :
docs/bundles.md. - byok :
docs/byok.md. - cli-reference :
docs/cli-reference.md. - connector-dsl :
docs/dsl.mdat the connector commit recorded in X09. - connector-local-code :
pkg/runview/localconnectors.goat that connector commit. - connector-cli-code :
cmd/iterion/connectors.goat that connector commit. - cloud-architecture :
docs/cloud-architecture.md. - cloud-git :
docs/adr/068-persist-run-diff-content-for-cloud-panels.md. - cloud-overview :
docs/cloud-overview.md. - cloud-rest-api :
docs/cloud-rest-api.md. - config-code :
pkg/configshare/service.go. - config-share :
docs/config-share.md. - credential-pool :
docs/credential-pool.md. - credpool-code :
pkg/credpool/broker.go. - current-state :
docs/current-state.md. - cursors :
docs/cursors.md. - delegation :
docs/delegation.md. - desktop :
docs/desktop.md. - desktop-projects-code :
cmd/iterion-desktop/bindings.go. - devbox :
docs/adr/017-devbox-first-bot-toolchain.md. - dispatcher :
docs/dispatcher.md. - dsl :
docs/dsl.md. - dsl-roadmap :
docs/adr/098-dsl-versioning-and-authoring-surface.md. - effort-code :
pkg/backend/model/effort.go. - event-primitives :
docs/adr/051-in-bot-event-driven-primitives.md. - file-secret-refresh :
docs/adr/069-mid-run-file-secret-refresh-for-sandboxed-runs.md. - forge-conversations :
docs/forge-conversations.md. - fork-code :
pkg/runview/fork.go. - github-board-sync :
docs/github-board-sync.md. - groups-iteration-subbots :
docs/groups-iteration-subbots.md. - human-in-the-loop :
docs/human-in-the-loop.md. - import :
docs/import.md. - improvement-ratchet :
docs/improvement-ratchet.md. - keychain-code :
cmd/iterion-desktop/keychain.go. - knowledge-code :
pkg/knowledge/iface.go. - license :
LICENSE. - mcp-server :
docs/mcp-server.md. - memory-and-knowledge :
docs/memory-and-knowledge.md. - memory-code :
pkg/backend/model/memory_tools.go. - native-tracker :
docs/native-tracker.md. - notifications :
docs/notifications.md. - oauth-forfait :
docs/oauth-forfait.md. - observability :
docs/observability.md. - observability-stack :
docs/observability/README.md. - outbound-callbacks :
docs/outbound-callbacks.md. - outbox-code :
pkg/trigger/effects_worker.go. - outcome-router :
docs/outcome-router.md. - pause-code :
pkg/runview/service_control.go. - permissions :
docs/permissions.md. - persisted-formats :
docs/persisted-formats.md. - pi-code :
pkg/backend/delegate/pi.go. - plugin-code :
pkg/plugin/manifest.go. - plugins :
docs/plugins.md. - post-mortem-shell :
docs/post-mortem-shell.md. - privacy-code :
pkg/backend/tool/privacy/register.go. - privacy_filter :
docs/privacy_filter.md. - quotas-and-limits :
docs/quotas-and-limits.md. - readme :
README.md. - recipes :
docs/recipes.md. - recovery-code :
pkg/runtime/recovery_dispatch.go. - repo-scope :
docs/repo-scope.md. - resume :
docs/resume.md. - review-merge-gate :
docs/review-merge-gate.md. - routers :
docs/routers.md. - run-file-editor :
docs/adr/016-in-run-worktree-file-editor.md. - run-recovery :
docs/adr/056-adaptive-run-level-recovery.md. - runner-code :
pkg/runner/loop.go. - sandbox :
docs/sandbox.md. - scheduling :
docs/scheduling.md. - scope-code :
pkg/knowledge/scope.go. - secrets-reference :
docs/secrets-reference.md. - sentry-feedback-loop :
docs/sentry-feedback-loop.md. - session-board :
docs/session-board.md. - session-persist :
docs/adr/089-session-persist.md. - settings-precedence :
docs/settings-precedence.md. - skills-library :
docs/skills-library.md. - sso :
docs/adr/035-per-org-sso-generic-oidc-plus-dedicated-connectors.md. - steering-code :
pkg/runview/steer.go. - subagents-code :
pkg/backend/model/executor_build_task.go. - subbot-reattach :
docs/adr/084-subbot-reattach-across-restarts.md. - supervise-code :
pkg/supervise/coordinator.go. - supervisors :
docs/supervisors.md. - tracker-code :
pkg/workspacetrack/tracker.go. - tracker-wiring-code :
pkg/runview/service_launch.go. - trigger-outbox :
docs/adr/094-trigger-effect-outbox.md. - trigger-spine :
docs/adr/046-event-driven-runs-trigger-spine.md. - ultracode :
docs/ultracode.md. - updater-code :
cmd/iterion-desktop/updater.go. - usage-caps :
docs/usage-caps.md. - vault-code :
pkg/backend/tool/privacy/vault.go. - verified-actions :
docs/adr/044-adaptive-recovery-for-deterministic-action-nodes.md. - verified-code :
pkg/backend/model/executor_verified_action.go. - visual-editor :
docs/visual-editor.md. - web-search :
docs/web-search.md. - webhooks :
docs/webhooks.md. - workspace-versioning :
docs/workspace-versioning.md. - worktree-finalization :
docs/adr/064-worktree-finalization-requires-delegated-authority.md. - worktree-pool :
docs/worktree-pool.md.