golemcore-bot

GolemCore Bot is a Java-based autonomous agent and framework that exposes MCP-enabled tool servers via standard input/output (stdio). It can run as a standalone agent (CLI or Telegram integration) or as a framework to build custom agents by composing skills, tools, and MCP integrations with routing and tiered LLM usage. The included tooling supports files-based skills (SKILL.md with YAML frontmatter), a browser tool (Playwright/Chromium) for web tasks, filesystem and shell operations, and sandboxed tool execution with safety rails. The MCP interface allows external tooling and services to attach as tool servers, enabling structured context sharing and controlled tool use across different models and workloads. When run via Docker as described, you can access a web dashboard at http://localhost:8080/dashboard and configure LLM providers, tools, and settings in the dashboard.

Prerequisites:

• Docker installed on your host (recommended) or a Java 25+ environment for a native run
• Access to an LLM provider (OpenAI, Anthropic, etc.) and corresponding API keys
• Git (optional, for cloning the repository)

Install via Docker (recommended):

1. Pull the published image: docker pull ghcr.io/alexk-dev/golemcore-bot:latest

2. Run the container (example): 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

3. Inspect logs to retrieve the temporary admin password on first run: docker logs -f golemcore-bot

4. Open the dashboard at http://localhost:8080/dashboard and configure LLM providers, keys, and tools in Settings.

If you prefer a local Java-based setup (non-Docker): ensure you have Java 25+ and Maven, then follow the repository's build instructions to run the application directly. The Docker path is the simplest way to run with default configurations and optional dashboard-based management.

Tips and notes:

• The container requires elevated privileges for the browser tool (Playwright/Chromium). The recommended flags (--shm-size, --cap-add=SYS_ADMIN) are applied in the Docker command.
• In Docker, ensure STORAGE_PATH is a mounted volume to persist configuration and sessions across restarts.
• Dashboard-based configuration stores sensitive keys in preferences/runtime-config.json. Keep backups of this file.
• The MCP integration allows external tool servers to communicate with GolemCore via stdio. You can expose custom tools or bridge other services through this interface.
• If you encounter issues starting the dashboard, check container logs for the initial admin password and verify that API keys and provider types are correctly set in Settings.
• The example uses the latest image tag; for production, pin to a specific version tag (e.g., :0.10.0) to ensure reproducibility.