submit: yai-agent-core

[Gi-Tuu]

X-Agent MCP Hackathon submission

• Submission directory: submissions/mcp-hackathon/yai-agent-core/
• Live API: https://yai-agent-core.onrender.com
• Review commit: fc7d3d6cd3362fcdec85dad195a9cd794e5b4616
• Source repository: https://github.com/Gi-Tuu/yai-agent-core

Submitter confirmation

• This PR changes one submission directory only.
• source/ contains the complete review source and no secrets or private data.
• RIGHTS.md identifies the submitter, third-party licenses, and archive authorization.
• The public health check reports the exact review commit.
• /.well-known/xagent-verification.json reports the same slug and commit.
• I have the right to submit this code, service, dependencies, and data for review.
• I understand that passing X-Agent review does not promise OKX acceptance, listing, traffic, or revenue.
• I understand that closing an unmerged PR withdraws the entry, and that accepted/rewarded source remains in X-Agent's official archive.

Reviewer notes

• Automated hard gates: package is 117 files / ~760 KiB (limits: 2000 files / 20 MiB total / 5 MiB per file); baseline secret scan clean; the review commit is publicly verifiable on GitHub; GET /health and GET /.well-known/xagent-verification.json both report the review commit; both responses are far below the 64 KiB cap. The official validator (npm run validate:submission -- --dir ...) passes offline; the online path was additionally verified end-to-end against the deployed service.
• What it is: YAI Agent Core — an embeddable, self-adaptive, in-process Agent Core (library-first; the host declares its capabilities and the core adapts at runtime, rather than shipping yet another standalone agent framework). The kernel itself has zero third-party runtime dependencies; model clients, the FastAPI battery, and the MCP/OpenAPI integrations are optional extras.
• Real capability evidence: the live demo host exposes native tools (notes search/count) and remote MCP tools (DeepWiki MCP server over Streamable HTTP). Real POST runs were verified end-to-end: a native-tool task returns the correct result with a complete event stream, and an MCP task performs a genuine remote MCP tool call and returns a correct grounded answer.
• Engineering: 125 offline tests (fake model / MockTransport / injected clocks — no network, no API keys), CI matrix (Ubuntu Python 3.11/3.12/3.13 + Windows 3.13), per-IP sliding-window rate limiting on POST runs (GET gate endpoints always reachable), opt-in history retention (message cap + TTL aligned to turn boundaries), and a keep-warm workflow so the free-tier instance stays reachable during the review window.
• Decision: pass

[github-actions]

Submission received — #56

Thank you for submitting to the current X-Agent MCP Hackathon. We have received your pull request.

This is a receipt only. It does not confirm eligibility, completeness, technical approval, a judging result, or an award. Passing automated checks or merging code for archival purposes does not, by itself, mean the entry has passed review or won a prize.

Please include the complete source for the submitted capability in this repository under submissions/mcp-hackathon/<slug>/source/, with dependency files, configuration examples, and reproducible setup/build instructions. An external repository link alone is not sufficient. The official copy is intended to remain available if the original repository later becomes unavailable. Do not include secrets, credentials, or private user data; clearly disclose external and private-service dependencies.

The automated checks report technical checks separately. A workflow failure is not a judging decision. Keep updates in this PR while submissions remain open, and keep the documented source version and deployment evidence consistent.

See the submission instructions. Final review results and awards will be announced separately through the official event channels.