agentwatch AGENTS.md

Repository instructions for agentwatch, a local tool that records activity from AI coding agents on one computer. It shows their sessions on a timeline with token and cost information, while reading logs stored on the machine rather than sending them to a cloud service.

In plain words
What is it for?
Use it to inspect agent sessions, token and cost usage, compaction events, and anomalies, or to query an agent's local history through its MCP server.
Why use it?
When several coding agents are running, it can be difficult to see what they have done or investigate unusual activity. These instructions explain how to install, verify, and use the local monitoring tool.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/mishanefedov/agentwatch/agents-md
Clone the repo
git clone --depth 1 https://github.com/mishanefedov/agentwatch

Made for: Codex, OpenCode.

Per session 1,142 This file is loaded in full into every session.
When invoked 1,142 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.01142 $0.01142
Opus 5 $0.00571 $0.00571
Sonnet 5 $0.00228 $0.00228
Haiku 4.5 $0.00114 $0.00114

Measured 2d ago against content hash 1c6a20f846b5, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agentwatch 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 2d 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.

AGENTS.md · 124 lines

How it starts

The opening of the file, as written. The whole thing — 124 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md

Guidance for AI agents. Two jobs — jump to yours:

  • Use agentwatch — you're an agent (or helping a user) who wants visibility into the AI agents running on this machine → Use agentwatch (install → verify → run; no account, no human needed)
  • Contribute to this repo — you're changing agentwatch's code → Working in this repo

Follows the AGENTS.md convention. Human contributors: see CONTRIBUTING.md.


Use agentwatch

agentwatch is local-only observability for every AI coding agent on one machine — Claude Code, Codex, Gemini CLI, Cursor, Hermes, OpenClaw — on a single timeline, with token + cost accounting, compaction + anomaly detection, and an MCP server agents can query their own history from. No cloud, no telemetry, no sign-in: it reads session logs already on disk. That means there's no auth step and nothing for a human to approve — you can onboard yourself end to end.

1. Install

npm i -g @misha_misha/agentwatch      # Node ≥ 20, macOS/Linux

The npm scope is @misha_misha (the bare name was taken); the binary on your PATH is agentwatch.

2. Verify

agentwatch doctor      # lists which agents are installed + whether events are captured

If an agent you use shows "not detected," that's the gap worth reporting — open an issue.

3. Run

agentwatch             # TUI live-tail + web UI at http://127.0.0.1:3456
agentwatch serve       # web UI only (remote boxes / headless)
agentwatch --help      # full flag reference

Press w in the TUI to open the web UI — that's where you drill into projects, sessions, token charts, call graphs, diff attribution, and replay.

No shell? Register the MCP server

For chat-only / IDE-embedded agents, expose agentwatch over MCP so you can query your own run history:

claude mcp add agentwatch -- npx -y @misha_misha/agentwatch mcp
# or edit ~/.claude.json / ~/.cursor/mcp.json by hand

Read the full file on GitHub · 124 lines

Changes

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.

  1. 2d ago First seen · 124 lines · 1,142 tokens per session scan A 1c6a20f846b5

Subscribe to this mod's changes

agentwatch AGENTS.md is an instructions file published in the GitHub repository mishanefedov/agentwatch (14 stars, last pushed 1mo ago), licensed MIT. It adds 1,142 tokens to every session, about $0.0057 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.