MCPOrb: The PDF for
AI-native knowledge delivery.

Package product docs, service playbooks, and expert knowledge into portable MCP Orbs that run locally inside AI tools.

One file. Everything inside.

Like a PDF, but for AI clients. Knowledge, index, configuration, and local Web UI — all in one portable Orb.

🗂

Portable

One Orb file contains your knowledge, search index, configuration, and runtime. Deliver it like a PDF — no cloud setup, no database, no API key required.

🖥

Local

Orbs run entirely on the user's machine. No data upload. No vendor lock-in. No compliance risk. Your knowledge stays where you put it.

🔍

Inspectable

Every Orb includes a local Web UI. See every document, chunk, source, and search result. No black box — open the browser and verify.

🤖

MCP-native

Works with Claude Desktop, Cursor, VS Code, and any MCP-compatible AI client. One config line and your knowledge is available as a search_knowledge tool.

From PDF to MCP tool in minutes.

MCPOrbBuilder turns your PDF or Markdown documents into a self-contained Orb. The Orb runs locally, exposes a search_knowledge MCP tool, and includes a local Web UI for inspection.

  • Build from PDF or Markdown with one command
  • Inspect the manifest, sections, and retrieval plan
  • Test queries before deploying
  • Launch the local Web UI to verify results
  • Copy the MCP config for Claude Desktop, Cursor, or VS Code
View on GitHub →
# Build an Orb from a PDF
mcporb build your-docs.pdf --name my-orb

# Inspect the result
mcporb inspect target/orbs/my-orb

# Test a query
mcporb test-query target/orbs/my-orb \
  "What is the onboarding process?"

# Launch Web UI
mcporb run target/orbs/my-orb --open

# Claude Desktop config
{
  "mcpServers": {
    "my-orb": {
      "command": "/path/to/my-orb",
      "args": ["--stdio-only"]
    }
  }
}

Built for teams that deliver knowledge.

MCPOrb is for B2B teams that need to package and deliver knowledge — not just store it.

B2B Software Teams

Turn product docs, API references, and migration guides into AI-ready delivery assets. Give customers an Orb they can use in their AI tools instead of searching static docs.

DevRel & Customer Success

Package onboarding guides, FAQ, and troubleshooting playbooks as Orbs. Reduce support load and improve developer self-service success rates.

Consulting & Service Teams

Deliver client-specific knowledge packs alongside project deliverables. Package methodology and expertise as a reusable AI assistant.

Internal AI Platform Teams

Distribute knowledge to internal teams without uploading sensitive data to external SaaS. Maintain data boundaries while enabling AI-assisted workflows.

Open runtime. Local execution.
Inspectable content.

Open Source Runtime is public

The MCPOrb Runtime is open source. Inspect the code, understand exactly what runs on your machine or your customer's machine. No black box.

Local No cloud dependency

Orbs run entirely locally. No API keys, no data upload, no vendor lock-in, no compliance risk. Your knowledge stays where you put it.

Auditable Every result is traceable

The local Web UI shows every chunk, source document, page number, and search score. You can verify every answer before it reaches your users.

RAG helps you answer questions.
MCPOrb helps you ship knowledge as a product asset — portable, inspectable, and deliverable to customers, partners, and teams.

Bring one document and five questions.

We're working with a small group of B2B software teams, consulting firms, and internal AI platform teams to build the first Orbs from real documents.

We'll run a free 30-minute Orb Fit Assessment to determine whether your content is a good fit, what the Orb would look like, and what success would mean for your team.

  • Free 30-minute Orb Fit Assessment
  • First Orb built and tested at no cost
  • Best practice recommendations for your content
  • Early access to MCPOrbBuilder
  • No commitment required

No spam. We'll respond within 2 business days.

✅ Request received! We'll be in touch within 2 business days.

In the meantime, check out the public demo on GitHub.