Evidence
Paste a tools/list result, its JSON-RPC response, or an array of tools.
Public preview Simulated payments · Fixed registry snapshot
Catch changes to tool schemas, Agent Cards, and payment terms before they reach your agents.
Paste a capture or start with an example.
Paste a tools/list result, its JSON-RPC response, or an array of tools.
Estimate the monthly cost of including this capture in an agent’s context.
Use the input rate you actually pay. If your client already compacts or caches input, modeled savings can differ.
Set one release policy for MCP, A2A, and x402 in your existing CI job.
Check which published MCP versions or declared transports changed before updating an integration.
Preview coverage and observed changes for free. The full brief adds version timelines, transport differences, and links back to the original records.
Checking the latest brief…
Recent active listings from the official MCP Registry.
Save a known-good baseline and run the same measurement in CI.
Capture the candidate in your existing test job. Set a compact-token budget and fail on changes when you need review.
Download sample capturenode src/cli.mjs baseline mcp known-good.json > baseline.json
node src/cli.mjs audit mcp candidate.json \
--baseline baseline.json --max-tokens 1000 --fail-on-change
Run from the AgentCaliper project. The audit returns JSON and a CI-friendly exit code.
Counts use o200k_base on the supplied or decoded JSON. Compacting removes whitespace outside strings and keeps numeric spellings. Model wrappers, caching, and tool selection can change what a provider bills.
Baselines store hashes, counts, and item labels. Comparisons identify changed definitions and selected contract fields. Export a baseline for each service and access profile you measure.
Findings cover selected structural checks. They describe the supplied capture. A live service, verified signatures, security, and payment settlement require separate evidence.