Local stdio MCP server for OrchestrateKit — evidence-backed playbooks for AI workflow design.
OrchestrateKIT has 19 tools with consistent naming (verb_noun pattern: list_, get_, compose_, plan_, etc.) and reasonable descriptions (100-250 chars typical). However, critical gaps exist: NO visible input schemas in provided source code, parameter descriptions are absent, and output schemas are undocumented. Tool definitions appear inferred from file names rather than explicitly registered with full schemas. The server implements tool annotations (toolAnnotations=true) but without schema visibility, cannot verify parameter types, constraints, or composition chains. Error handling and recovery guidance are not evident in sampled code.
Compose a custom workflow route by specifying source and target components, with automatic path finding and edge validation.
Get a detailed explanation of what a component does, its inputs, outputs, and typical use cases.
Export a build brief summarizing the current registry state, component counts, and deployment readiness.
Retrieve detailed information about a specific component by ID, including its description, inputs, outputs, and related edges.
Retrieve detailed information about a specific edge by ID, including its severity, test status, and test references.
Retrieve a complete playbook by ID, including its goal, route, clarifying questions, and runtime recommendations.
Retrieve relevant documentation and guidance for a given component, edge, or workflow context.
NO input schemas visible in source code. All 19 tools scored 0 for schema dimension. Cannot verify parameter types, constraints, enums, or composition chains.
Parameter descriptions missing. Sampled code (toolSafetyRules.ts) shows tool definitions but no parameter-level documentation. LLMs cannot infer what 'type', 'status', 'search_term', 'source', 'target', 'severity' mean without explicit descriptions.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 35 | 2026-07-28+ | v2 |
Retrieve detailed information about a specific workflow route by ID, including its components, edges, and inline evidence summaries.
Get technology stack recommendations based on workflow requirements and constraints.
Returns the server name, version, a summary of loaded registry entities (components, edges, stacks, routes, playbooks), and a build fingerprint. The `build.stale` field is true when the dist/ registry is outdated vs the source — run `pnpm build` to fix. The `safe_to_demo` field is true only when the build is fresh, the registry meets the published component/edge/route/playbook count floors, AND `build.content_fingerprint` matches the pinned expected release fingerprint; `demo_blockers` lists any reasons it is not. `matches_expected_release` distinguishes the two ways a build can fail trust: a build can have counts that clear every floor ("counts compatible") while still not being the exact published registry snapshot ("matching release" is false) — for example a hosted build serving an older or newer, unreleased registry. Use this to confirm the MCP server is running and the registry is fresh.
List all components in the registry with optional filtering by type, status, or search term.
List all edges (connections) in the registry with optional filtering by source, target, relation type, or test status.
List all known workflow routes in the registry with optional filtering by trigger type, components, or automation level.
Generate a detailed workflow plan from a high-level goal, including route composition, clarifying questions, and runtime recommendations.
Get architecture recommendations based on workflow requirements, scale, and operational constraints.
Record user feedback about a planning session for continuous improvement of recommendations.
Replay a previous planning session using a stored passport to reproduce the same recommendations and decisions.
Review a workflow design against safety rules, architecture patterns, and best practices, returning findings and recommendations.
Validate a candidate playbook against registry rules, safety constraints, and best practices.
Output schemas undocumented. Tool descriptions state what they return (e.g. 'detailed information', 'recommendations', 'findings') but do not specify field names, types, or structure. Agents cannot plan downstream calls or extract required IDs.
Error handling and recovery guidance absent. Sampled code shows tool safety rules but no error classification (retryable vs fatal), no actionable error messages, and no recovery paths. Agents cannot self-correct on failures.
Tool composition chains not documented. No evidence that get_route returns IDs needed by compose_workflow_route, or that plan_workflow output feeds into validate_playbook_candidate. Broken chains force discovery detours.