MCP server for Exponential - connect Claude to your projects, actions, and goals
Exponential MCP has 4 tools with clear naming (verb_noun pattern) and reasonable descriptions (100-250 chars). All tools have input schemas with type definitions and parameter descriptions. However, output schemas are not documented, LLMs cannot predict response structure. Parameter descriptions are present but minimal (20-50 chars), lacking format constraints, ranges, or dependency hints. No error handling guidance. Missing tool annotations (readOnlyHint/destructiveHint). The update_project tool is complex with 10 optional parameters but lacks clear guidance on mutual exclusivity or side effects of wholesale replacement (goalIds/keyResultIds). Overall solid foundation but falls short of production-grade polish.
List actions/tasks, optionally filtered by project or status. This is a flat list with no date filtering — to answer "what should I work on today" or "what am I behind on", use get_todays_actions instead.
Fetch one project with its relations: the objectives and key results it is linked to, DRI, team, dates, description, and linked meetings. get_projects returns none of the OKR links, so this is the tool for "what is this project driving?". It does NOT include the project's tasks — use get_actions with projectId for those. Accepts a project ID, a slug, or the slug-id form from app URLs.
List all projects in Exponential. Returns project names, status, and progress.
Update a project: rename, change status or priority, edit the description, set dates, or re-link it to objectives / key results. Only the fields you pass are changed. goalIds and keyResultIds REPLACE the existing links wholesale — read get_project first and pass the full set you want to keep. An empty keyResultIds clears the key-result links; an empty goalIds is ignored by the server.
Output schemas not documented. LLMs cannot predict response structure (fields, types, nesting). Increases hallucination risk and forces exploratory calls.
update_project lacks clarity on parameter relationships. goalIds/keyResultIds 'REPLACE the existing set wholesale', but what if the user only wants to add one? No guidance on read-before-write pattern or partial updates.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Agents cannot distinguish safe reads from destructive writes without parsing descriptions.
Parameter descriptions are minimal (20-50 chars). Missing format constraints, ranges, and dependency hints. E.g., 'status' enum values are documented but no guidance on valid state transitions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
No error handling guidance. If update_project fails due to invalid goalIds or circular OKR links, LLM has no recovery path.