Add AGENTS.md: MCP + HTTP usage contract for agents

[Khamel83]

AGENTS.md stub committed. Needs: example MCP call patterns for 'topic research' vs 'URL extraction', example POST /api/search payload per mode, and link validation to khamel83/maya docs/CONTEXT-CONTRACT.md.

[Khamel83]

Research Findings

The objective of this task is to draft a policy document, AGENTS.md, which establishes a structured contract for agents interacting with MCP systems using HTTP methods. This document will clarify how agents should properly engage in 'topic research' versus 'URL extraction' tasks, defining the expected payloads and validation protocols for different usage scenarios.

The first section highlights the importance of creating clear operational guidelines that align with MayaKit’s Contrat Stack documentation, ensuring consistency across interactions. Researchers should focus on implementing standardized MCP message patterns for each intent—such as retrieving relevant topics or parsing content from web URLs—while maintaining traceable contracts.

Key findings include the necessity of documenting typical HTTP verb behaviors (GET for topic retrieval, POST for content extraction) and ensuring payload formatting adheres to API specifications. Further, payload validation must reference external guidelines, such as those outlined in the Contrat Stack's CONTEXT-CONTRACT.md.

Important open questions arise around:

• Defining counterparty responsibilities for topic accuracy.
• Identifying edge cases for URL extraction failures.
• Establishing versioning strategies for evolving HTTP contracts.

To address these, resources such as the Khamel83 documentation and the Maya API reference should be reviewed closely.

The essential references include:

• MayaKit Contrat Stack documentation
• Contrat Stack API guide
• APIs MCP Request & Response Examples

In summary, this initiative aims to create a robust, extensible framework for agent-MCP communication by formalizing expected usage patterns and contractual obligations. Proper implementation will empower agents to operate reliably and transparently within Maya workflows.

---

Generated by Clio research lane (Lane B)

[Khamel83]

Research complete. Findings re-ingested into Clio for next-action routing.