agentleFS
systemind-teamsystemind-team/AICOO-Skills/skills/connect-local-agent/SKILL.md

connect-local-agent

Skill35 starsChanged 4 months agoMITView on GitHub

Use this skill when the user wants to make THIS machine's coding agent reachable for live agent-to-agent collaboration — start the Aicoo local-agent bridge so paired peers can send requests to their running Claude Code / Codex, with per-tool owner approval. Triggers on: 'connect my local agent', 'start the bridge', 'make my agent reachable', 'set up local agent', 'local agent collaboration', 'let others reach my agent', 'run the Aicoo bridge', 'connect this device', or after the in-app 'Set up your local agent' prompt. NOTE: this is the RECEIVING side (be reachable); to CONTACT someone else's agent use talk-to-agent.

  • Reads credentials
  • Installs packages

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    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
  1. Very specific:

    If AICOO_API_KEY is missing, get one at https://www.aicoo.io/settings/api-keys (or run onboarding).

    line 24
  2. Distinctive:

    If it's still empty a heartbeat later, set it manually as a fallback:

    line 97
  3. Very specific:

    Keep this process running: If it dies when the session ends, run the same nohup … npm run bridge … line in a normal terminal, or set up a …

    line 110

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Connect Local Agent — start the collaboration bridge

You set up and start the Aicoo local-agent bridge on the user's machine. The bridge registers this device with Aicoo and holds a live Claude Code / Codex session, so another person's agent can send it a request and get an answer back — with every tool call gated on the owner's approval (a message conveys intent, not authority).

This is the reachable side. The bridge is a long-lived process — it must keep running.

Prerequisites

  • AICOO_API_KEY environment variable set (aicoo_sk_live_...)
  • Claude Code CLI ≥ 2.1.211 (the bridge drives it as the local runtime)
  • git + Node.js
echo "${AICOO_API_KEY:+API key set (${#AICOO_API_KEY} chars)}" || echo "NO AICOO_API_KEY"
claude --version   # need ≥ 2.1.211; install/upgrade: npm i -g @anthropic-ai/claude-code

If AICOO_API_KEY is missing, get one at https://www.aicoo.io/settings/api-keys (or run onboarding).


Step 1: Get the bridge

if [ ! -d aicoo-local-agent ]; then
  git clone https://github.com/Aicoo-Team/aicoo-local-agent.git
fi
cd aicoo-local-agent && git pull --ff-only 2>/dev/null; npm install

The bridge is the open-source Aicoo Local Agent (Apache-2.0) — anyone can clone it. It bundles both a self-hostable reference control plane and the hosted-Aicoo transport used below.


Step 2: Start the bridge (background, keep alive)

Server: production https://www.aicoo.io, or a preview the user names (e.g. https://yourcoo.ai) via CCD_SERVER_URL.

SERVER="${CCD_SERVER_URL:-https://www.aicoo.io}"
SPOOL="me.spool"   # remember this file — the asker side needs it for its reply route

nohup env \
  CCD_AICOO=1 \
  CCD_SERVER_URL="$SERVER" \
  CCD_TOKEN="$AICOO_API_KEY" \
  npm run bridge -- --adapter claude-code --spool "$SPOOL" \
  > bridge.log 2>&1 &
echo "bridge starting (pid $!) against $SERVER — logging to bridge.log"

CCD_AICOO=1 selects the Aicoo transport; CCD_TOKEN reuses the API key (an OAuth access token works too); the bridge registers its endpoint + a managed session and heartbeats. deviceId is auto-generated and persisted next to the spool. It does not auto-set the default route — Step 4.


Step 3: Confirm it registered

sleep 3
head -40 bridge.log

Success = a JSON block with an endpointId, no auth/connection errors. Report the endpointId and confirm it is heartbeating (every ~20s).

Common failures:

  • 401 / unauthorized despite a valid key → the apex host did a cross-origin redirect (aicoo.io → www.aicoo.io) that dropped the Authorization header. Use the www. host in CCD_SERVER_URL (e.g. https://www.aicoo.io). Also check the key isn't expired / belongs to the account on that server (a preview may use a different database).
  • connection refused / timeout → check CCD_SERVER_URL and network.
  • claude not found → add --claude-path "$(which claude)" to the bridge command.
  • No conversation found on restart → a prior crash left a stale session. The adapter now auto-recovers (starts a fresh session); if it still surfaces, clear the stale spool session state (keep the .device-id file) and restart.

Step 4: Verify the default route (usually automatic)

The default route is what makes you reachable — it maps "someone targeting your person_default_runtime" to this specific endpoint + session. The bridge now auto-sets it from its heartbeat loop (within ~20s of start). Just verify:

CCD_AICOO=1 CCD_SERVER_URL="$SERVER" CCD_TOKEN="$AICOO_API_KEY" \
  npm run ccd -- default-route get

If it's still empty a heartbeat later, set it manually as a fallback:

CCD_AICOO=1 CCD_SERVER_URL="$SERVER" CCD_TOKEN="$AICOO_API_KEY" \
  npm run ccd -- default-route set --spool me.spool

Step 5: What's now possible

  • The agent is reachable. In the Aicoo app, anyone the user has paired with (agent access granted) can hit Collaborate and send a request; the user gets an Accept popup, then per-tool Allow/Deny prompts.
  • Keep this process running. If it dies when the session ends, run the same nohup … npm run bridge … line in a normal terminal, or set up a launch agent for persistence.
  • Tools: by default the receiver is text-only (safe deny-all). A permissioned mode (每 tool call → owner Allow/Deny) is available opt-in and still being finalized — until then, first-test questions should be answerable without tools (e.g. "reply with PONG").

Quick reference

StepWhat happens
PrereqAICOO_API_KEY (or OAuth token) + Claude Code CLI present
1Clone/update the bridge, npm ci
2Start bridge in background (www. host!) — deviceId auto
3Confirm endpointId + heartbeat in bridge.log
4Verify the default route (auto-set by the bridge; default-route get)
5Reachable — paired peers can send requests (owner-approved)

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.