DeepORG (formerly GraphJin) — one AI agent for your whole organization: databases, APIs, files, and code over MCP

DeepORG — One AI Agent for Your Whole Organization

Apache 2.0NPM PackageDocker PullsDiscord ChatGoDoc

DeepORG was formerly GraphJin. The CLI (graphjin), the Go module, the npm package and the Docker images keep the graphjin name for now.

DeepORG is an AI agent that answers questions and makes changes across thesystems your company already runs: databases, warehouses, APIs, files andsource code. Ask it in plain words; it finds the right tables, has the databasedo the work, and answers with the data behind it. It can only see and changewhat your access rules allow, and by default a change waits for your approval.

Why it's different

No hand-built tools. Most company agents are wired up one tool at a time:a "get customer" tool here, a "list invoices" tool there, each written andmaintained by hand. They break when a schema changes, and every new questionor new system means more tools. DeepORG gives the agent a searchable map ofeverything you connect, and the agent writes its own queries against it. Theengine checks each query against your access rules and runs it as one efficientdatabase query. Connecting a new database or API is a few lines of config, nota new set of tools.

Proven on a public benchmark. DeepORG, the Organizational AgentBenchmark scores AI models on realorganizational work against a live system: answers, changes, alerts, follow-upsand questions that span several sources. The best published run scores 97 of100 with zero unsafe actions, and every score links to the exact test and buildthat produced it.

Turn your organization into a training environment. The same machinerygenerates checked tasks from your own systems and serves them as gradedepisodes, so you can measure an agent, or train one with reinforcementlearning, on your organization instead of a toy.Agent environment

Underneath is GraphJin, the open-source engine that compiles each query into asingle database query. It is also a GraphQL API server and a Go library in itsown right.

Install

curl -fsSL https://deeporg.io/install.sh | bash   # macOS and Linux
npm install -g graphjin                              # any platform
brew install dosco/graphjin/graphjin                 # macOS

Windows: use npm install -g graphjin, or download the Windows archive fromthe releases page. Linux packagesare on the same page, and Docker is docker pull dosco/graphjin.

Try it

graphjin serve --demo

This starts a demo SaaS company (accounts, subscriptions, invoices, supporttickets) with seeded data. Put a model key in ./.env (OPENAI_API_KEY,ANTHROPIC_API_KEY or GOOGLE_API_KEY) and ask it something:

curl -sS localhost:8083/api/v1/agent \
  -H 'content-type: application/json' \
  -d '{"instruction": "Which account is most at risk of churning?"}'

Or chat with it at http://localhost:8083/agent. More demos:Demos.

Use it from Claude, Codex or any MCP client

graphjin mcp add claude http://localhost:8083

Other clients: MCP. Hosted servers with sign-in:MCP OAuth.

Learn more

Get started Quick start · Your first query
The agent How it works · Conversations · Watches · Workflows
Connect your systems Databases · APIs · Files · Source code
Run it in production Production · Access rules · Config reference
GraphQL and Go Query language · Go examples
Measure agents Benchmark · Agent environment

Works with PostgreSQL, MySQL, SQLite, MongoDB, Oracle, SQL Server, Snowflake,BigQuery, Cassandra, S3, GCS and local files.

Community

Discord · Twitter @dosco

License

Apache 2.0

MCP Server · Populars

MCP Server · New

    dosco

    DeepORG — One AI Agent for Your Whole Organization

    DeepORG (formerly GraphJin) — one AI agent for your whole organization: databases, APIs, files, and code over MCP

    Community dosco
    scumble-app

    Scumble

    Free, open-source desktop editor for AI inpainting: select, prompt, get a colour-matched layer. FLUX 3 Image, FLUX.2, GPT Image, Nano Banana, Seedream and Qwen through your API key or your own ComfyUI. Layers, PSD export, MCP.

    Community scumble-app
    langmail

    caldav-mcp

    🗓️ A CalDAV Model Context Protocol (MCP) server to expose calendar operations as tools for AI assistants.

    Community langmail
    mozilla

    Firefox MCP

    Model Context Protocol (MCP) server that lets AI agents browse the web and debug web pages with Firefox

    Community mozilla
    maestro-press

    Crow

    Modular, agentic framework and MCP platform you self-host. Build and run your own AI agents, connect Claude Code, Claude Desktop, or Cursor as a native MCP connector, self-host your apps, and share over encrypted P2P — on hardware you own, with local or cloud models. Open source.

    Community maestro-press