Public assessment

keepgoing-dev/mcp-server (@keepgoingdev/mcp-server)

keepgoingdev-mcp-server · v0.1.0 · scanned Aug 3, 2026

What changed in the harness

Selection accuracy 95→95, token cost up 1%, unconfirmed writes 0%→0%.

Category breakdown

Where the score comes from.

Earned points across the four signals Gradable measures. Safety and Legibility are scored out of 30; Economics and Discoverability are scored out of 20.

01Safety

0.0 / 30

0.0 out of 30

02Legibility

25.4 / 30

25.4 out of 30

03Economics

19.7 / 20

19.7 out of 20

04Discoverability

17.2 / 20

17.2 out of 20

Highest-impact fix

Estimated gain +30 points

Add explicit identity and permission preflight tools

Expose machine-readable principal/tenant confirmation and a non-mutating permission check so agents can verify both before destructive actions.

Description evidence

Defects and rewrites.

7 defects found across the exposed tool descriptions. Suggested rewrites make purpose, inputs, boundaries, and returns easier for an agent to understand.

Tool	Defect types	Suggested rewrite
`get_momentum`	no_return_description	Get current developer momentum and return a briefing containing the last checkpoint, next step, blockers, and branch context. Use this to understand where the developer left off. Pass tier or model to control detail level.
`get_reentry_briefing`	no_return_description	Get a synthesized re-entry briefing that helps a developer understand where they left off. Returns the briefing with focus, recent activity, and suggested next steps. Pass tier or model to control detail level.
`get_current_task`	no_return_description	Return a bird's eye view of all active Claude sessions, showing what each session is working on, which branch it is on, and when it last did something. Useful when running multiple parallel sessions across worktrees.
`save_checkpoint`	no_return_description	Save a development checkpoint and return confirmation that it was recorded. Call this after completing a task or meaningful piece of work, not just at end of session. Each checkpoint helps the next session (or developer) pick up exactly where you left off.
`get_context_snapshot`	no_return_description	Return a compact context snapshot containing what you were doing, what is next, and momentum. Use this for quick orientation without a full briefing.
`get_whats_hot`	no_return_description	Return a summary of activity across all registered projects, sorted by momentum, showing what the developer is working on across their entire portfolio.
`setup_project`	no_return_description	Set up KeepGoing hooks and instructions and return a status message describing what was configured. Use scope "user" for global setup (all projects) or "project" for per-project setup.

Selection evidence

Confusable tool pairs.

6 pairs where similar names or overlapping descriptions may send an agent toward the wrong tool.

Tool A	Tool B	Confidence	Why they collide
`get_momentum`	`get_context_snapshot`	high	Both describe near-identical function: get_momentum reports 'last checkpoint, next step, blockers' and says 'understand where the developer left off', while get_context_snapshot reports 'what you were doing, what is next, and momentum' for 'quick orientation'. A task like 'where did I leave off' or 'what's next' maps plausibly to either.
`get_momentum`	`get_reentry_briefing`	high	Both take identical tier/model schema and serve the same purpose: get_momentum 'understand where the developer left off' and get_reentry_briefing 'understand where they left off' with focus/recent activity/next steps. A re-entry or orientation task could select either.
`get_current_task`	`get_context_snapshot`	medium	The name get_current_task strongly suggests 'my current task', but its description is a multi-session overview; get_context_snapshot covers personal 'what you were doing'. A task phrased 'what am I currently working on' could be routed to either.
`get_current_task`	`get_whats_hot`	medium	Both are activity-overview tools: get_current_task summarizes active Claude sessions/branches, get_whats_hot summarizes activity across projects. A vague 'give me an overview of what's going on' task could select either depending on session vs project framing.
`get_session_history`	`get_reentry_briefing`	medium	A task like 'what did I work on recently' fits both the chronological session-checkpoint list and the synthesized re-entry briefing with 'recent activity'; the agent must infer whether raw history or a synthesized summary is wanted.
`get_momentum`	`get_current_task`	medium	For a task like 'what am I working on now', get_momentum (last checkpoint, next step, blockers) returns the personal state while get_current_task's name matches 'current task' but its description is the parallel-session overview; either could be chosen.

Compare the field

One score is useful.
The evidence makes it actionable.

Back to the leaderboard