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/brainos-hq/brain-os/agents-mdgit clone --depth 1 https://github.com/brainOS-HQ/brain-osWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/brainos-hq/brain-os/agents-md)<a href="https://agentmods.dev/instructions/brainos-hq/brain-os/agents-md"><img src="https://agentmods.dev/badge/instructions/brainos-hq/brain-os/agents-md.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.03348 | $0.03348 |
| Opus 5 | $0.01674 | $0.01674 |
| Sonnet 5 | $0.00670 | $0.00670 |
| Haiku 4.5 | $0.00335 | $0.00335 |
Grade A, and why
brain-os 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 5d 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 — 215 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Agent operating instructions for the Brain OS repository.
This file is the canonical source of truth for how AI agents (Claude Code, GitHub Copilot, Cursor, Zed, Windsurf, Codex, any MCP-compatible client) should behave when working in this repo. Tool-specific files (CLAUDE.md, .github/copilot-instructions.md, .cursor/rules/brain-os.mdc) are thin pointers to this document.
Context
This repo IS Brain OS — an MCP server for persistent operational state across AI agents. State lives in .brain/. The MCP server is built from src/, published to npm as brain-os, and exposes tools prefixed mcp__brain-os__* when connected as a client.
When asked about project state, priorities, decisions, plans, blockers, or focus, you are answering questions about an operating product, not a codebase. Use the MCP tools. Do not grep source code.
The hard rule
For ANY question about project state, priorities, decisions, patterns, focus, or what to work on, use the Brain OS MCP tools as the primary data source. Never read src/tools/focus-get.ts (or similar) to answer "what should I focus on" — that's the implementation of the tool, not the answer. Call the tool itself.
If you read source code as your first action on a state question, you are in generic-coding-assistant mode. Stop and re-route through MCP.
Tool routing (in order)
| # | Tool | When to call |
|---|---|---|
| 1 | mcp__brain-os__entity_read(entity_id?) |
First call for ordinary project-state questions. Omit entity_id to list all entities. |
| 2 | mcp__brain-os__plan_read(entity_id) |
Get active step + progress for an entity. |
| 3 | mcp__brain-os__context_resolve(user_message?, files_touched?, explicit_entity_id?) |
Resolve context before focus_get / decision_check when the target entity is not already known. For "what should I work on" from inside a workspace, call this before focus_get. |
| 4 | mcp__brain-os__project_evidence_scan(root_path, entity_id?) |
Read-only repo evidence (STATE.md, FLAGS*, HANDOFF*, ROADMAP/PLAN/TODO, git activity, dirty files). Call AFTER context_resolve and BEFORE focus_get when a workspace path is known. NOT a router — never use it to pick the project. |
| 5 | mcp__brain-os__focus_get(entity_id?, constraints?) |
Prioritized recommendations. Pass resolved entity_id to scope to one project; omit only for explicit global focus or unresolved context. |
| 6 | mcp__brain-os__semantic_recall(query, source_kind?) |
Fuzzy search when you don't know the entity ID or want cross-decision / pattern / session context. |
| 7 | mcp__brain-os__decision_check(proposed_action, entity_id?) |
Call before any action that might contradict an active decision. Returns clear / caution / conflict, plus review_triggered — decisions whose invalidate_if condition the action matches (premise may have changed; surface for review, don't enforce). |
| 8 | mcp__brain-os__pattern_detect() |
Surface current behavioral patterns. |
| 9 | mcp__brain-os__entity_update, plan_update, decision_log, decision_refresh, memory_commit |
Mutating tools. Use when writing state back. |
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.
- 5d ago First seen · 215 lines · 3,348 tokens per session scan A cf15478e7b04
brain-os AGENTS.md is an instructions file published in the GitHub repository brainOS-HQ/brain-os (6 stars, last pushed 11d ago), licensed MIT. It adds 3,348 tokens to every session, about $0.0167 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 2026-08-31.
Other instructions, from other repositories
turbo_quant_memory AGENTS.md
AGENTS.md instructions for Lexus2016/turbo_quant_memory, covering agents.md, project identity, memory usage policy, session continuity (start-of-session and compaction) and accessing project secrets (phase 9).
turbo_quant_memory CLAUDE.md
Claude Code instructions for Lexus2016/turbo_quant_memory, covering claude.md, tqmemory workflow, practical rule and secrets vault (phase 9).
plur CLAUDE.md
Claude Code instructions for plur-ai/plur, covering claude.md, what is plur, development, package dependency and version bumps.
engraphis AGENTS.md
AGENTS.md instructions for Coding-Dev-Tools/engraphis, covering agents.md — engraphis, internal subagent delegation, 0. read this first — two architectures live in one package, 1. commands and ── unified dashboard + memory inspector ──.
brigade AGENTS.md
AGENTS.md instructions for escoffier-labs/brigade, covering agents.md, definition of done, rules, orientation and memory handoff.
sverklo CLAUDE.md
Instructions for sverklo/sverklo, covering sverklo development, using sverklo tools, project structure and build & test.