- Category
- AI Agents
- Rank
- No. 1988Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- openclaw
- GitHub
- 24 stars
- Latest release
- workspace-latest
- Date
About
An OpenClaw experimental Q&A interface project.
What it does
Ask Molty answers OpenClaw documentation questions through an authenticated web service. It retrieves matching documentation, source excerpts, and GitHub discussions, mounts a small read-only workspace, then lets an OpenAI model search that evidence before composing a cited answer.
Why it's ranked here
The design has unusually clear evidence priorities: documentation defines intended behavior, source resolves implementation details, and GitHub threads supply discussion or status. Candidate selection limits per-request context, while restricted tools reduce exposure. Its narrow OpenClaw focus makes the value specific rather than broad.
What's good
Retrieval degrades thoughtfully. Stored artifacts can survive a public documentation outage, the full documentation corpus backs up a failed search index, and source or GitHub context remains usable when documentation disappears. The simulated shell permits familiar inspection commands without network access, writes, redirects, pipes, or process execution.
Tradeoffs
Deployment depends on Cloudflare Workers, an OpenAI API key, ClawHub authentication, and externally published retrieval artifacts. GitHub evidence comes from a SQLite mirror rather than live GitHub. User messages are capped at 2,000 characters, tool work stops after four rounds, and only selected workspace files reach each request.
How to use it well
Use it when OpenClaw users need grounded installation, configuration, debugging, or extension guidance with citations. It fits beside published documentation as an evidence-aware support layer. Maintain the exported indexes and smoke checks before deployment. It does not replace general-purpose Q&A, live GitHub monitoring, documentation publishing, or unrestricted repository exploration.
Technical notes+
src/worker.ts implements the Cloudflare Worker routes, authenticated chat flow, four-round tool loop, restricted shell emulator, and OpenAI streaming path. src/retrieval.ts loads R2 or remote indexes, ranks records, falls back to the documentation corpus, and mounts at most 32 workspace files. src/prompt.ts defines evidence precedence and citation policy. scripts/export-workspace.ts exports flat Markdown artifacts and JSONL indexes from documentation, source, and Gitcrawl SQLite data. scripts/smoke.ts checks artifact size, auth routing, R2 preference, corpus fallback, and partial retrieval during documentation outages. package.json keeps the package private and exposes npm scripts for checking, exporting, local Worker development, and Wrangler deployment. wrangler.toml binds the R2 bucket and configures production routes.
Observed
- License
- MIT License
- Primary language
- TypeScript
- Packaging and install surface
- Private ECMAScript module package installed with npm; operational scripts run through npm.
- Interface
- Authenticated HTTP documentation chat service with session, sign-in, callback, chat, and artifact routes.
- Platform support
- Cloudflare Workers deployment with an R2 bucket binding and Wrangler configuration.
- Quality checks
- The check script runs strict type checking, linting, formatting validation, workspace export testing, and smoke checks.
Read from README.md, package.json, src/auth.ts, src/types.ts, src/prompt.ts, src/worker.ts, src/retrieval.ts, scripts/smoke.ts, scripts/export-workspace.ts, LICENSE, AGENTS.md, .oxfmtrc.json, wrangler.toml, .oxlintrc.json, tsconfig.base.json.
What it can do
Answer user questions
Text-based questions → Text-based answers
Process natural language queries
Natural language questions → Structured responses
Provide interactive Q&A sessions
Sequential user questions → Conversational responses
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.