MiddleClaw is a lightweight AI agent that acts as a middle-man between you and OpenClaw. It communicates, queries, and manages your OpenClaw installation through natural conversation — relaying information and executing actions, but only when you say so.
MiddleClaw bridges you and OpenClaw with the power of local AI — all under your explicit control.
Describe what you need in plain language. MiddleClaw translates your intent into OpenClaw operations and relays the results back, acting as a seamless middle-man.
Detects your OpenClaw installation and understands its configuration, services, and logs. Communicates with OpenClaw using context-aware bridging.
Run parallel communication sessions with persistent browser history. Bridge multiple OpenClaw queries simultaneously.
Commands adapt automatically to Linux, macOS, or Windows. No manual translation needed.
Run .sh, .bash, .bat, .cmd, and .ps1 scripts from configured directories with sandboxed execution.
Full dark mode support with a clean theme toggle. Easy on the eyes during those late-night bridging sessions.
Configure every parameter through a clean graphical interface. Model selection, path permissions, port settings, and more — no config file editing required.
MiddleClaw keeps you in control at every stage. Nothing happens without your explicit approval.
Tell MiddleClaw what you want to communicate to OpenClaw in plain language. Check status? Query configuration? Relay a command? Just describe it like you would to a colleague.
MiddleClaw proposes the commands it needs to bridge your request to OpenClaw and explains what each one does. You review and approve every single action before it runs.
systemctl status openclaw && curl -s http://localhost:8080/health | jq .
MiddleClaw relays the results from OpenClaw, interprets them for you, and suggests next steps if needed. Backups are created automatically before any file modifications.
Every design decision prioritizes your system's safety. MiddleClaw has guardrails at every level.
Every file read, command execution, script run, and file write requires your explicit approval. Nothing happens in the background.
Dangerous operations like rm -rf, mkfs, and shutdown are blocked at the pattern level. They can't even be proposed.
Before modifying any file, MiddleClaw creates a backup in .middleclaw-backups/. Roll back instantly if anything goes wrong.
Individual commands timeout after 30 seconds. Scripts timeout after 60 seconds. Runaway processes are automatically terminated.
File operations are sandboxed to configured directories. MiddleClaw can't access paths you haven't explicitly allowed.
Runs on your machine via Ollama. Cloud models are supported if you choose to enable them, but never required.
MiddleClaw requires Node.js 18+ and Ollama. Install with a single command or clone and configure manually.