Tiny mcp-agent -> Assay runtime accounting sample (token_summary)

[Rul1an]

Hi all — we put together a very small sample on our side to test one honest runtime-accounting seam from mcp-agent, and I wanted to share that rather than open with a broad integration pitch.

The sample is here:
https://github.com/Rul1an/assay/tree/main/examples/mcp-agent-token-evidence

It stays deliberately narrow. We treat token_summary from the documented token-counter path as the current best first seam, keep model_breakdown optional, treat cost_estimate_usd only as an upstream estimate, and do not treat token accounting, workflow outcomes, or MCP protocol behavior as Assay truth.

The sample is docs-backed and frozen on purpose. The runnable token-counter examples in mcp-agent are great, but for this first pass we wanted one bounded external-consumer artifact without dragging provider setup or runtime dependency into the claim.

This is not an MCP protocol evidence proposal. It is just a tiny runtime-accounting handoff sample inside an MCP-native framework.

If an external evidence consumer wants the smallest honest mcp-agent surface for bounded runtime accounting evidence, is token_summary roughly the right seam, or is there a thinner exported surface you’d rather point them at?

Happy to keep this small and follow your lead on the right boundary.