Q
AGENTS.md vs CLAUDE.md?
AGENTS.md is the cross-tool standard; CLAUDE.md is Claude Code's equivalent. Keep both thin or converge on AGENTS.md with tool-specific files for extras.
Glossary · AGENTS.md
AGENTS.md.
One instruction file every agent reads.
Direct answer
AGENTS.md is the cross-tool standard for repo-wide agent instructions (donated to the Linux Foundation's Agentic AI Foundation, Dec 2025): conventions, stack, test/build commands, architectural invariants — read by Cursor, Codex, OpenCode, Anvaya, and more. Measured impact: ~29% lower median runtime and ~17% fewer output tokens on PR-sized tasks. Rules: only what the agent can't infer from code, short, versioned in PRs, pruned ruthlessly; nested files merge with specific-wins.
In Anvaya
How We Implement It.
01AGENTS.mdAnvaya reads AGENTS.md as standing Write-layer context and handles everything it can't: accumulated why, bug history, and patterns that would bloat a hand file.
Questions
Asked About AGENTS.md.
Q
What belongs in it?
Stack + versions, test/lint/build commands, non-obvious conventions, hard constraints. Not: codebase tours the agent can read itself.
Q
Monorepo strategy?
Nested files: root holds universal truths, packages hold specifics, specific-wins merging. Scope rules by glob so agents load only their package.
Stop Starting From Zero.
One binary. 11+9 Rust crates. 545 tests. Hand-written HNSW index. Three transport modes. Four providers, Ollama, Anthropic, OpenAI, Siemens. Zero API keys required to start. Mind remembers everything after the first session.