scout
AGENTS.md11 starsChanged 10 days agoApache-2.0View on GitHub
This repository is the Scout core. It holds the runtime, the protocol, the harness adapters, the scout CLI and MCP server, the TUI, scoutd, and the host side of phone pairing. The Mac, iOS, Android, and web apps and the hosted services live in a separate repository. Do not add app or UI source here. docs/public-source-boundary.md draws the boundary. Core code must not import app source. Behavior that a client needs moves down into a package here first. - The…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,542 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository4 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- release-trainSkill
- scoutSkill
The file
We split this AGENTS.md into the rules it gives an agent
7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 44
Use Bun 1.4.4 or newer.
- Very specific: line 64
Run the check for each package you change: bun run --cwd packages/<name> check.
- Distinctive: line 67
Run the narrowest affected tests while you work.
- Very specific: line 69
Before you hand off a cross-package change: bun run check and bun run test:unit.
- Very specific: line 80
Commit subjects use gitmoji: 🐛 Fix session route resolution, 📝 Clarify runtime ownership.
- Distinctive: line 82
Do not add AI co-author trailers or "Generated with" lines.
- Very specific: line 88
Do not run the release, ship, publish, or bump scripts unless a maintainer asks you to.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Scout core
This repository is the Scout core. It holds the runtime, the protocol, the
harness adapters, the scout CLI and MCP server, the TUI, scoutd, and the
host side of phone pairing. The Mac, iOS, Android, and web apps and the hosted
services live in a separate repository. Do not add app or UI source here.
docs/public-source-boundary.md draws the
boundary.
Layout
| Path | Holds |
packages/protocol | @openscout/protocol: contracts, identities, routing metadata, record types |
packages/runtime | @openscout/runtime: the runtime, persistence, service custody, mesh, delivery |
packages/cli | @openscout/scout: the CLI and MCP server, published with the runtime and pairing bundled |
packages/pairing | Host side of phone pairing: encrypted bridge, controller, phone RPC |
packages/agent-sessions | Harness adapters and session extraction |
packages/session-trace | Headless live-session trace model |
packages/mesh-extension | HTTP adapter for the default MCP core tools |
packages/scout-tailscale | Python client for owner-approved guest access over Tailscale |
crates/scoutd | Native daemon for the base service: supervisor, probes, repo scan and diff |
crates/scout-tui | Terminal UI |
crates/openscout-repo-service | Compatibility wrapper over the scoutd repo service |
docs/agent/ | Dense specs for agents. Start at docs/agent/README.agent.md |
Core code must not import app source. Behavior that a client needs moves down into a package here first.
Rules
- The runtime is the only writer of Scout records. Clients and adapters send commands through its API.
- Harness transcripts are observed source material. Do not import them as Scout messages.
- One target is a DM. A group needs an explicit channel. Broadcast is opt-in.
- Route from target metadata. A message body is payload, not routing.
- Mesh means reachability and coordination. It does not mean exactly-once delivery, consensus, or transcript replication.
- A harness or session mismatch must fail with an actionable diagnostic. It must not hang.
- Scout assumes trusted local users and agents. Do not claim enterprise, compliance, or multi-tenant readiness.
- Use Bun 1.4.4 or newer.
Test against your own instance
A checkout runs as its own Scout instance, with its own state, ports, and service label. Test there. The installed runtime on port 43110 belongs to the person whose machine it is.
bun run runtime:brokerstarts thedevinstance: runtime 24400, web 24410, pairing 24420–24422. A linked worktree getsdev-<folder>and its own ports.scout --instance dev <command>talks to that instance. Plainscouttalks to the installed runtime.scout instance listshows the claimed instances and their ports.- You must not stop, restart, or repoint the installed runtime
(
app.openscout) unless you are asked to.
docs/dev-instances.md has the details.
Checks
- Run the check for each package you change:
bun run --cwd packages/<name> check. Protocol, runtime, pairing, cli, agent-sessions, session-trace, and mesh-extension have one. - Run the narrowest affected tests while you work.
- New or changed process spawning or shell execution:
bun run sync-exec:fence. - Before you hand off a cross-package change:
bun run checkandbun run test:unit. - CI also runs
bun run check:core(no app imports) andbun run check:presentation(public repository presentation). - Rust:
bun run scoutd:check,bun run scoutd:test,bun run scout:tui:check,bun run repo-service:check. bun run docs:generateregeneratesllms.txtanddocs.jsonfromdocs/. This file is written by hand. Dewey does not generate it.
Commits and pull requests
- Commit subjects use gitmoji:
🐛 Fix session route resolution,📝 Clarify runtime ownership. - Do not add AI co-author trailers or "Generated with" lines.
- Push a topic branch to your fork and open the pull request against
openscout-app/scout. CONTRIBUTING.md lists what the description must contain. - This repository is public. Do not commit local paths, host names, credentials, or private source.
- Do not run the
release,ship,publish, orbumpscripts unless a maintainer asks you to.
Read next
- docs/agent/README.agent.md: records, the first-run ladder, and the rules for routing and runtimes.
- docs/architecture.md: the data model and identity.
- CONTRIBUTING.md: setup and pull requests.
<!-- scout:begin -->Scout
- Work for another agent, or on another machine, goes through Scout:
scout ask(MCPask). - Another machine:
scout ask --on <host> --project <path on that host> --harness <claude|codex>;scout hostslists machines and their projects. Don't ssh in to launch agents. - Syntax:
scout ask --help,scout help --detail. Added byscout setup agents-md;scout setup agents-md --removetakes it out.
<!-- scout:end -->Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.