GolemCore Bot

Agent Platform for AI-Native Companies — build, run, extend, and inspect channel-connected agents with skills, plugins, MCP, memory, and Hive.

---

What it is

GolemCore Bot is the agent platform for AI-native companies building on the GolemCore ecosystem.

Use it to run channel-connected agents, extend them with skills, plugins, and MCP servers, and operate them with memory, delayed follow-ups, trace inspection, and Hive-connected control flows.

Use it in two ways:

1. Run agents — chat in the dashboard, Telegram, or webhook-driven flows, then inspect execution, memory, and trace output.
2. Extend the runtime — add skills, install plugins, connect MCP servers, configure routing tiers, and attach Hive-based control and inspection.

Why it is different

• Plugin-native runtime — plugins can contribute tools, channels, voice providers, and RAG ingestion instead of living outside the agent runtime.
• MCP-ready by design — skills can launch MCP servers, and the runtime now includes MCP catalog and discovery flows.
• Built for long-running work — delayed follow-ups, progressive memory orchestration, and recovery paths help agents continue past a single turn.
• Hive-connected control surface — connect agents to Hive for control commands, lifecycle signals, inspection, and host-level coordination.
• Inspectable execution — traces, waterfall views, and exported snapshots make agent behavior easier to debug and operate.

---

Quick start

What you need

• Docker
• At least one LLM provider API key

Run the published image

docker pull ghcr.io/alexk-dev/golemcore-bot:latest

docker run -d \
  --name golemcore-bot \
  --shm-size=256m \
  --cap-add=SYS_ADMIN \
  -e STORAGE_PATH=/app/workspace \
  -e TOOLS_WORKSPACE=/app/sandbox \
  -v golemcore-bot-data:/app/workspace \
  -v golemcore-bot-sandbox:/app/sandbox \
  -p 8080:8080 \
  --restart unless-stopped \
  ghcr.io/alexk-dev/golemcore-bot:latest

docker logs -f golemcore-bot

# Open http://localhost:8080/dashboard
# On first start, check logs for the temporary admin password.

Open http://localhost:8080/dashboard, sign in with the temporary admin password from the logs, then configure your LLM providers in Settings.

Why the extra Docker flags?

• --shm-size=256m and --cap-add=SYS_ADMIN are needed for the Browser tool (Playwright/Chromium) in containers. See Configuration → Browser Tool.

Optional next steps

• Enable Telegram in Settings for channel-based chat.
• Install plugins for browser, search, mail, weather, voice, or RAG-backed workflows.
• Connect MCP-backed skills and tool servers.
• Join Hive for control-plane coordination and inspection.

Need a local build, Compose setup, or production deployment path? See Quick Start and Deployment.

---

First-run setup

1. Open the dashboard.
2. Add at least one LLM provider key in Settings.
3. Verify the storage and sandbox volumes are mounted so sessions, skills, and runtime config persist.
4. Optionally enable Telegram, install plugins, connect MCP-backed skills, or join Hive.

For runtime config details, storage layout, and browser or sandbox notes, see Configuration.

---

Documentation

Documentation site: https://docs.golemcore.me/

The source docs below are kept in this repo for local and offline reference:

1. Quick Start
2. Skills
3. Tools
4. MCP Integration
5. Model Routing
6. Memory + RAG
7. Auto Mode + Delayed Actions Design
8. Hive Integration
9. Webhooks
10. Deployment
11. Dashboard

---

Contributing / Support / License

• Contributing: see CONTRIBUTING.md (workflow, quality gates, tests).
• FAQ: FAQ.md.
• Issues: https://github.com/alexk-dev/golemcore-bot/issues
• License: Apache 2.0 — see LICENSE and NOTICE.

---

⭐ Star this repo if you find it useful!

Made with ☕ and 🤖