← 返回
开发者工具 中文

IDE Agent Kit

Filesystem message bus and webhook relay for multi-agent IDE coordination. Use when agents need to share events, poll Ant Farm rooms, receive GitHub/GitLab w...
文件系统消息总线和 webhook 中继,用于多智能体 IDE 协调。适用于共享事件、轮询 Ant Farm 房间以及接收 GitHub/GitLab 等 webhook。
thinkoffapp thinkoffapp 来源
开发者工具 clawhub v0.4.1 1 版本 99902.5 Key: 无需
★ 0
Stars
📥 1,025
下载
💾 8
安装
1
版本
#ack-only filtering#developer-tools#file-based notifications#ide#latest#local-first#message rate limits#room polling

概述

IDE Agent Kit

Connect your IDE coding agents into real-time teams through OpenClaw. Filesystem-based message bus, room polling, automation rules, and multi-model agent coordination. Zero dependencies.

Security Model

This skill operates in two tiers:

Core (local-only, no credentials needed):

  • Local filesystem queue and receipt log — agents read/write files in the working directory
  • init, receipt tail, memory (local backend), keepalive — no network, no secrets
  • serve binds to 127.0.0.1 only by default — receives webhooks, writes to local queue

Advanced (requires explicit opt-in and credentials):

  • sessions, gateway — talk to an OpenClaw gateway (requires openclaw.token in config)
  • poll — connects to Ant Farm rooms (requires --api-key flag)
  • emit, hooks create — POST data to external URLs you specify
  • tmux run, exec — execute shell commands (restricted to an allowlist in config)

No advanced features activate without explicit configuration. The default init config has empty credential fields and a minimal command allowlist.

Network behavior

CommandOutbound connectionsInbound connections
---------------------------------------------------
init, receipt tail, memory (local), keepaliveNoneNone
serveNonelocalhost:8787 only (configurable)
pollAnt Farm API (HTTPS)None
sessions, gatewayOpenClaw gateway (localhost by default)None
emitUser-specified URLNone
hooks createUser-specified webhook URLNone

Command execution

tmux run and exec only run commands listed in tmux.allow in your config. Default allowlist: npm test, npm run build, pytest, git status, git diff. Commands not on the list are rejected.

exec adds an approval flow: commands go through exec request → human/agent exec resolve before running.

Quick Start

npm install -g ide-agent-kit
ide-agent-kit init --ide claude-code

Creates a local ide-agent-kit.json config. All credential fields are blank. Nothing connects to any server until you configure it.

Connectivity Modes

Four modes that compose freely. Only mode 1 is active by default.

1. Local Filesystem Bus (default)

Agents on the same machine communicate through a shared queue directory and receipt log. No network, no server, no API keys.

  • Queue: ./ide-agent-queue.jsonl
  • Receipts: ./ide-agent-receipts.jsonl

2. Webhook Relay Server (optional)

Receives inbound webhooks from GitHub/GitLab and writes them to the local event queue.

ide-agent-kit serve [--config <path>]

Binds to 127.0.0.1:8787 by default. Set github.webhook_secret in config to verify signatures. Does not make outbound connections.

3. Ant Farm Room Polling (optional)

Connects to Ant Farm rooms for cross-machine coordination.

ide-agent-kit poll --rooms <room1,room2> --api-key <key> --handle <@handle> [--interval <sec>]

Requires: --api-key flag (Ant Farm API key). Rate-limited, default 120s interval.

4. GitHub Events (optional)

When serve is running, point a GitHub webhook at your relay URL. Translates PR/issue/CI events into local queue events.

Requires: github.webhook_secret in config to verify inbound signatures.

Commands

Core (local-only, no credentials)

CommandDescription
----------------------
`init [--ide ] [--profile low-friction>]`Generate starter config
receipt tail [--n ]Print last N receipts
watch [--config ]Watch event queue, nudge IDE session on new events
serve [--config ]Start webhook relay server (localhost only)
`memory list\get\set\search`Manage agent memory (local file backend)
`keepalive start\stop\status`Prevent macOS sleep for remote sessions

Advanced (requires credentials or explicit config)

CommandRequiresDescription
--------------------------------
sessions send --agent --message openclaw.tokenSend message to agent via gateway
sessions spawn --task openclaw.tokenSpawn a new agent session
`sessions list\history\status`openclaw.tokenQuery sessions
`gateway trigger\health\agents`openclaw.tokenGateway operations
poll --rooms --api-key --handle Ant Farm API keyPoll rooms for messages
emit --to --json None (user specifies target)POST event JSON to a URL
hooks create --webhook-url None (user specifies target)Create webhook forwarder
tmux run --cmd Allowlisted commands onlyRun command in tmux, capture receipt
`exec request\resolve\list`Allowlisted commands onlyExecution approval workflow
`cron add\list\remove\run\status`openclaw.tokenScheduled task management

Configuration

Generated by ide-agent-kit init. All credential fields default to empty.

FieldPurposeDefault
-------------------------
listen.hostWebhook server bind address127.0.0.1
listen.portWebhook server port8787
tmux.allowAllowlisted shell commands[npm test, npm run build, pytest, git status, git diff]
openclaw.tokenGateway auth (advanced commands)empty
github.webhook_secretVerify GitHub webhooksempty

Data Access

PathAccessPurpose
-----------------------
ide-agent-receipts.jsonlappendAudit log of all agent actions
ide-agent-queue.jsonlread/writeEvent queue
ide-agent-kit.jsonreadRuntime configuration (may contain secrets)
memory/read/writeLocal agent memory files

Source & Verification

  • npm: https://www.npmjs.com/package/ide-agent-kit
  • Source: https://github.com/ThinkOffApp/ide-agent-kit
  • Maintainer: petruspennanen (npm), ThinkOffApp (GitHub)
  • License: AGPL-3.0-only

The npm package contains no install scripts (preinstall/postinstall). All code is plain ESM JavaScript. Verify with npm pack --dry-run before installing.

版本历史

共 1 个版本

  • v0.4.1 当前
    2026-03-29 15:59 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

ai-agent

self-improving agent

pskoett
捕获经验教训、错误及修正内容,以实现持续改进。适用于以下场景:(1)命令或操作意外失败;(2)用户纠正Claude(如“不,那不对……”“实际上……”);(3)用户请求的功能不存在;(4)外部API或工具出现故障;(5)Claude发现自身
★ 4,090 📥 818,778
ai-agent

ontology

oswalpalash
类型化知识图谱,用于结构化智能体记忆与可组合技能。适用于以下场景:创建/查询实体(人物、项目、任务、事件、文档)、关联相关对象、强制执行约束、将多步操作规划为图谱变换,或当技能需要共享状态时。触发关键词包括"记住""我知道关于什么""将X链
★ 725 📥 245,389
ai-agent

Skill Vetter

spclaudehome
AI智能体技能安全预审工具。安装ClawdHub、GitHub等来源技能前,检查风险信号、权限范围及可疑模式。
★ 1,233 📥 268,724