Toolslip for agents and platforms
Toolslip exposes cached public-alpha verdicts for 50 observed MCP servers as static JSON, so agent runtimes, registries, crawlers, and human integrators can inspect pre-connect risk signals before opening a session. The current surface is read-only and cached: it is not a live scanner, certification, audit, or guarantee.
Current recommended_action counts: 27 allow, 6 allow with caution, 17 human review.
There are two audiences for this surface: agents and crawlers consume static files; humans and integrators decide whether to wire Toolslip into an agent runtime, registry, or gateway.
Available now
How an agent can use the static surface
- Fetch /llms.txt to discover the public surface.
- Inspect /api/v1/stats.json for coverage and action counts.
- Fetch featured verdicts for the current public-alpha slice.
- Follow per-server JSON links from the server index when a runtime needs a specific cached record.
- Treat recommended_action as a pre-connect signal, not permission to bypass human approval.
What to measure
Without adding backend tracking, demand can be read from distinct static paths: /for-agents/ for human integrator interest, /llms.txt for crawler discovery, /api/v1/agent-surface.json for machine-surface discovery, stats JSON for lightweight machine interest, featured verdicts JSON for deeper machine interest, and server index or per-server JSON requests for bulk or specific-server interest.
Planned, not live yet
There is no live arbitrary query_server(url) today. Real-time lookup, arbitrary URL queries, paid fresh probes, and structured machine-readable decision objects are planned, not available.
Building an agent runtime?
This contact path is for human builders and operators, not autonomous agents. Tell us what you are building, which integration surface you need, which MCP servers matter, and how to contact you.