Turn AI-authored documents into business-ready deliverables.
You're working with an AI assistant. Together, you're producing real work — design documents, technical specs, integration plans, meeting summaries. The output is Markdown, because that's the native authoring format for AI: structured, version-controllable, rich with embedded code and diagrams.
Now share that work with your colleagues.
You can't send them a .md file. You can't ask them to install a Markdown viewer. You need something they can read in a browser, download as a PDF, or open in Word — today, without friction.
Jeeves Server gives you a secure, polished window into the machine where your AI assistant lives. It turns the raw output of AI collaboration into something your team can actually use:
- Browse and view any file on the server — Markdown renders beautifully with table of contents, syntax-highlighted code, and Mermaid diagrams
- Share securely — generate expiring links for external recipients, no login required
- Export instantly — one-click PDF and DOCX downloads, perfectly rendered
- Stay in control — Google OAuth for your team, scoped API keys for integrations, all zero-trust
You don't author documents here. That's what your AI assistant is for. Jeeves Server is the publishing layer — the bridge between the laboratory and the boardroom.
Jeeves Server also includes a webhook gateway — a durable, schema-validated event processing pipeline. External services (Notion, GitHub, CI/CD systems, anything that sends webhooks) can POST to your server, and Jeeves will:
- Validate the payload against JSON Schema rules
- Transform the body using JsonMap (extract just the fields you need)
- Queue matched events in a durable JSONL queue that survives restarts
- Dispatch to shell commands with the transformed payload on stdin
This makes Jeeves Server a natural integration hub for the same AI-assisted workflow: your assistant writes automation scripts, Jeeves Server receives the triggers, and the scripts execute on the same machine where everything else lives. No Lambda functions, no cloud queues — just a webhook URL and a command.
See the Event Gateway guide for setup and configuration.
Markdown is the native document format for AI collaboration:
- AI assistants read and write it natively — no format translation, no lossy conversion
- It supports everything business documents need: headings, tables, code blocks, diagrams, links
- It's plain text — version-controllable, diffable, mergeable
- It's the format your assistant already thinks in
The gap has always been the last mile: getting Markdown into the hands of people who don't know what Markdown is. Jeeves Server closes that gap.
- File Browser — Navigate drives and directories through a modern React UI
- Markdown Rendering — Prose with TOC sidebar, adjustable reading width, dark/light themes
- PDF & DOCX Export — One-click, perfectly rendered, business-ready
- Code Highlighting — Syntax highlighting with copy buttons
- SVG, Mermaid & PlantUML Diagrams — Rendered inline with pan/zoom; PlantUML uses a fallback pipeline (local jar → private servers → community server)
- Embedded Diagrams in Markdown —
mermaidandplantumlfenced code blocks render as inline SVGs (GitHub convention) - Secure Sharing — Expiring links with HMAC signatures, scoped access
- Event Gateway — Webhook receiver with JSON Schema validation and durable queue
- Zero CDN — All assets served locally, no external dependencies
- Setup & Configuration — Installation, auth modes (Google OAuth & key-based), config structure, and getting running
- Insiders, Outsiders & Sharing — The access model, how share links work, key derivation, expiry, and rotation
- Exporting & Downloads — PDF, DOCX, and ZIP export, Puppeteer setup, troubleshooting, and what affects rendered output
- Event Gateway — Webhook receiving, JSON Schema matching, body mapping, durable queue processing, and monitoring
- Deployment — Running as a service (Windows & Linux), reverse proxy setup, HTTPS, Google OAuth for production, and updates
- API & Integration — Programmatic access, endpoint reference, Windows path conversion, generating share links, and notes for AI assistants
The 📖 button in the header is itself a live external share link — it points to this very README—in the working directory on a live production Jeeves Server—with depth=2 (so you can follow links to the guides) and dirs=false (no directory browsing). If you're reading this in Jeeves Server, you're seeing the deep share feature in action. If you're reading this on GitHub, click here to see for yourself.
Jeeves Server runs on Windows and Linux.
- Windows: File browser auto-discovers drive letters (A-Z). Chrome path defaults to Program Files.
- Linux: File browser uses configurable
rootsinjeeves.config.ts(e.g.{ home: '/home', projects: '/opt/projects' }). Chromium path is typically/usr/bin/chromium-browser.
Both platforms are tested in CI (GitHub Actions on Ubuntu, Node 20 + 22).
git clone https://github.com/karmaniverous/jeeves-server.git
cd jeeves-server
npm install
cp jeeves.config.template.ts jeeves.config.ts # Configure
npm run build
cd client && npx vite build --outDir ../dist/client && cd ..
node dist/server.jsFull requirements and architecture: .stan/system/stan.requirements.md
MIT