Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/atomic-mail/atomic-mail-agentic/agents-mdgit clone --depth 1 https://github.com/Atomic-Mail/atomic-mail-agenticWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.02198 | $0.02198 |
| Opus 5 | $0.01099 | $0.01099 |
| Sonnet 5 | $0.00440 | $0.00440 |
| Haiku 4.5 | $0.00220 | $0.00220 |
Grade A, and why
atomic-mail-agentic AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 138 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Guidance for AI coding agents working in this repository.
What this repo is
Atomic Mail Agentic ships MCP and CLI clients for Atomic Mail—a hosted email inbox for AI agents. Everything built in this repo — the local stdio MCP server, the AgentSkill CLI, the LangChain tools, the Python parity layer — exposes three operations only: register, jmap_request, and help.
That is the client surface, not the product surface. The hosted remote MCP server at mcp.atomicmail.ai is a different artifact, built server-side, and exposes 11 tools (read_inbox, read_message, search_messages, send_email, reply_to_message, list_agents, search, fetch, run_preset, jmap_request, help). Do not "fix" a doc that mentions those tools by trimming it back to three.
Primary implementation: TypeScript on Deno (ts/). Python (py/) mirrors core behavior for parity; keep it aligned with shared assets.
Agent host → MCP or CLI → ts/src/lib → auth.atomicmail.ai → api.atomicmail.ai (JMAP)
Sibling repo: the server
Everything to the right of ts/src/lib in that diagram is built somewhere else.
The backend that implements auth.atomicmail.ai, api.atomicmail.ai, and
mcp.atomicmail.ai lives in a separate, private repository (agentic-mail).
It is not a submodule and it is not vendored here.
If you have it checked out locally, it may be a sibling of this repo (try
../agentic-mail or ../../agentic-mail); otherwise ask the operator.
If you do not have access to it — which is the normal case — then:
- Treat every server behavior as a fixed API contract, documented in
docs/on this side. Readdocs/rest-auth.md,docs/jmap.md,docs/oauth.md, anddocs/mcp-remote.mdbefore assuming what a request or response looks like. - Do not infer the contract from client code alone. The client is one implementation of it; it can be wrong, and it can be behind.
- When a change would require the server to change too, stop and say so rather than guessing. Shipping a client that sends a field no endpoint accepts is worse than not shipping: mocked tests pass, and the failure surfaces in production. Say which endpoint needs to change and hand it back to the operator.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 138 lines · 2,198 tokens per session scan A c5509baccf41
atomic-mail-agentic AGENTS.md is an instructions file published in the GitHub repository Atomic-Mail/atomic-mail-agentic (256 stars, last pushed 7d ago), licensed MIT. It adds 2,198 tokens to every session, about $0.0110 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
fastmail-mcp-server copilot-instructions.md
Copilot instructions for Jordonh18/fastmail-mcp-server, covering copilot instructions for fastmail mcp server, project overview, architecture, directory structure and tool groups registered in server.ts.
Hermes3D AGENTS.md
Instructions for iamlukethedev/Hermes3D, covering agent instructions, development environment instructions, service overview, running the app and lint, typecheck, and tests.
cli AGENTS.md
Instructions for nylas/cli, covering ai coding agent guidelines, quick reference: shared helpers, quick reference: adding a new feature and quick reference: credential storage.
better-email-mcp CLAUDE.md
Instructions for n24q02m/better-email-mcp, covering claude.md - better-email-mcp, commands, setup, lint & type check and fix.
olkcli CLAUDE.md
Instructions for rlrghb/olkcli, covering claude.md, what is this project?, quick reference, architecture and key patterns.
imap-mcp-server CLAUDE.md
Instructions for nikolausm/imap-mcp-server, covering claude.md and claude-specific notes.