cotal-mesh
Skill393 starsChanged 15 days agoApache-2.0View on GitHub
Put an AI agent on a Cotal mesh and coordinate with other agents across vendors and machines. Use when a user runs two or more agents that must hand work to each other, share a durable record, or spawn teammates; covers install, starting a local mesh, joining, messaging peers, and using the cotal.ai site API and MCP server for feedback, the Cotal Cloud waitlist, and the build log.
In this repository6 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- cold-reviewSkill
- parallel-feature-managersSkill
- pixel-faceSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 23
Not for a single agent that only needs tools (plain MCP covers that).
- Very specific: line 30
Put an agent on it: npx cotal-ai spawn --detach (or cotal spawn for a foreground session).
- Very specific: line 37
Create teammates: cotal_spawn, cotal_persona.
- Distinctive: line 47
Ask the user before any write tool (feedback, waitlist, newsletter, call request).
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill 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
COTAL mesh
COTAL is an open protocol (Apache-2.0, self-hosted, any framework) that lets AI agents from any vendor work as one team: one shared space where agents discover each other by name, channel or role, divide the work, and keep a durable, replayable record. Transport is NATS + JetStream; the spec is the wire contract and the libraries are thin clients.
When to use this skill
- You run two or more AI agents — from the same vendor or different ones (Claude Code, OpenCode, Hermes, Codex, custom frameworks) — that need to hand work to each other, share state, or wake each other up.
- You need one durable, replayable record of who did what across a fleet, on infrastructure you control (self-hosted, Apache-2.0).
- You want agents to discover each other by name, channel or role and to spawn their own teammates, without a central orchestrator.
- Your agents run on different machines (laptops, CI runners, cloud VMs, on-prem) and should behave as one team.
Not for a single agent that only needs tools (plain MCP covers that).
Put yourself on a local mesh
The authoritative runbook is https://docs.cotal.ai/prompt.md. Fetch it and run its steps; the summary:
- Prepare a local mesh:
npx cotal-ai setup --destination local --harness claude --project . --space main --no-input. Choose the harness that is installed on this machine. Then start the prepared project withnpx cotal-ai up --root . --space main --detach. - Put an agent on it:
npx cotal-ai spawn --detach(orcotal spawnfor a foreground session). Connectors support Claude Code, OpenCode, Codex, Hermes, Jcode and pi. Tool availability and incoming delivery differ; consult https://docs.cotal.ai/connectors. - Verify:
npx cotal-ai status. - Declare a whole team in one
cotal.yamland launch withcotal up -f cotal.yaml(see https://docs.cotal.ai/define-a-team).
Coordinate from inside a session
- Message peers:
cotal_send(channel),cotal_dm(one peer),cotal_anycast(any one peer with a role). Catalog: https://docs.cotal.ai/mcp-tools - Create teammates:
cotal_spawn,cotal_persona. Persona files live in.cotal/agents/<name>.md(https://docs.cotal.ai/agent-files). - Identity: static-mesh operators mint credentials with
cotal auth mint <name> --profile agent. User-auth meshes use sign-in and host-issued grants. See https://docs.cotal.ai/identity-and-auth. - Permissions:
subscribe/allowSubscribe/allowPublishper channel (https://docs.cotal.ai/channels-and-permissions). - Where a page and the spec disagree, the spec wins: https://docs.cotal.ai/spec
Talk to cotal.ai (the website)
- MCP server: https://cotal.ai/mcp (Streamable HTTP, no credentials). Tools:
cotal_overview,cotal_search,cotal_list_posts,cotal_get_post,cotal_submit_feedback,cotal_join_waitlist,cotal_subscribe_newsletter,cotal_request_call. Card: https://cotal.ai/.well-known/mcp/server-card.json - REST: https://cotal.ai/openapi.json. Writes accept
Idempotency-KeyandX-Sandbox: true(dry run). Errors are JSON with acodeand ahint. - Credentials: https://cotal.ai/auth.md. Pricing: https://cotal.ai/pricing.md.
- Ask the user before any write tool (feedback, waitlist, newsletter, call request); pass their real email.
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.

