Skip to content
Get packs

Guide · Open source

How to use DocsGPT: install, build agents and embed them

Updated · written with AI from the project's own documentation and checked against our catalogue data

DocsGPT is an open-source platform that turns your documents, websites and connected apps into AI agents that answer questions and cite their sources. It is aimed at teams that want an internal assistant over their own documents, and at developers who want to embed a documentation chatbot on a product site. After this guide you will know how to install it, what the main pieces are, and how to use agents, sources and the API.

DocsGPTarc53/DocsGPT · ★ 18.4k · MIT

Quick start

  1. 1Install Docker. DocsGPT's listed prerequisite is Docker. Install it before running the installer.
  2. 2Run the installer. On macOS or Linux, run the one-command installer. On Windows, run the PowerShell command. The installer runs docsgpt up.
  3. 3Choose a model. During docsgpt up, pick who should reach DocsGPT and which model to use, cloud or local.
  4. 4Open the app. A local install opens at http://localhost:7091 in your browser.
  5. 5Add knowledge. Upload your documents or connect a service such as GitHub or Google Drive.
  6. 6Create and test an agent. Set the agent's prompt, knowledge and tools, then ask it questions and check the cited sources.
  7. 7Back up and upgrade. Use docsgpt backup before running docsgpt upgrade.

What DocsGPT does

DocsGPT takes your documents, sites and connected apps and lets you build AI agents on top of them. Each agent can have its own prompt, knowledge, tools and model. Answers cite the documents they came from.

The project is MIT licensed and written in Python. The README says it can run entirely in your own environment, using cloud or local models.

  • Agents and a deep research mode for multi-step answers
  • Visual workflows that chain agents, conditions and state
  • Knowledge from PDFs, Office files, web pages, audio and connected services
  • Answers with sources
  • Any model: OpenAI, Anthropic, Google, Groq, OpenRouter, or local models through Ollama, vLLM and other OpenAI-compatible servers

Install DocsGPT on macOS, Linux or Windows

The README lists Docker as the prerequisite. On macOS or Linux, run the one-command installer. The installer checks for Docker, installs the docsgpt command and runs docsgpt up, which asks who should reach DocsGPT and which model to use.

On Windows, use PowerShell with the command below. A local install opens at http://localhost:7091.

  • Make sure Docker is installed before you start
  • Local install address: http://localhost:7091
  • Prefer Docker Compose, pip, Kubernetes or an air-gapped install? The README links to a deployment guide that covers those options
curl -fsSL https://docs.ac/install | bash irm https://docs.ac/install.ps1 | iex

Manage the stack with the docsgpt command

After the install, the docsgpt command manages the stack. Use it to check status, follow logs, upgrade, back up data and stop the stack. Stopping keeps your data and settings.

Run a backup before you upgrade. The README describes backup as covering the database and uploaded data.

  • docsgpt status shows version, address and health
  • docsgpt logs follows the logs
  • docsgpt upgrade upgrades and restarts on the new version
  • docsgpt backup backs up the database and uploaded data
  • docsgpt down stops the stack; data and settings stay
docsgpt status docsgpt logs docsgpt upgrade docsgpt backup docsgpt down

Add sources and build an agent

Start by adding knowledge: upload files or connect a service. The README names Google Drive, SharePoint, Confluence, GitHub and S3 as sources that can be kept in sync. Then create an agent with a prompt, the knowledge it should use, and tools if needed.

Once the agent works, you can use it in the web app, embed it on a site, or call it through the API.

  • Add PDFs, Office files or web pages as knowledge
  • Connect a service such as GitHub and choose a sync schedule
  • Give the agent a prompt, knowledge, tools and a model
  • Build a visual workflow when you need conditions or several agents working together

Embed DocsGPT or call its API

You can put an agent on a website with a script tag or the React package. DocsGPT also offers an OpenAI-compatible API, so any OpenAI SDK can point at its /v1 path. The Agent API and webhooks are separate options, and an MCP server lets clients such as Claude or Cursor search an agent's knowledge.

Read the security checklist in the docs before you expose DocsGPT beyond your own machine.

  • Chat and search widget: script tag or React package
  • OpenAI-compatible API at /v1
  • Agent API and webhooks
  • MCP server for MCP clients

Use local models with Ollama or vLLM

If you want models to run on your own hardware, the README says you can run local models through Ollama, vLLM and other OpenAI-compatible servers. Embeddings can also run locally, and the README mentions air-gapped deployments.

The README does not give model recommendations or quality benchmarks, so test answers against your own documents before relying on them.

  • Choose a cloud provider or a local model during setup
  • Test answers on your own documents, since the README makes no accuracy claims

What people use it for

  • An internal assistant that answers staff questions from company wikis and documents
  • A documentation chatbot embedded on a product support page
  • A private question-answering tool for data that must stay in your own infrastructure
  • A local-model setup using Ollama or vLLM
  • An MCP server that lets a coding assistant search your agent's knowledge

Frequently asked questions

Is DocsGPT free?

The code is MIT licensed, and the README says everything, including SSO, teams and quotas, is MIT licensed. The hosted DocsGPT Cloud at app.docsgpt.cloud is a separate service, and this source does not list its prices.

What do I need to install DocsGPT?

The README lists Docker as the prerequisite. The one-command installer checks for Docker and then installs the docsgpt command.

How do I install DocsGPT on Windows?

Run the PowerShell command irm https://docs.ac/install.ps1 | iex. The installer then runs docsgpt up, which asks about access and the model to use.

Can DocsGPT run with local models?

Yes. The README says you can run local models through Ollama, vLLM and other OpenAI-compatible servers, and that it can run entirely in your own environment.

How is DocsGPT different from AnythingLLM or Open WebUI?

DocsGPT focuses on agents, visual workflows and grounded answers over your documents, with an embeddable widget and API. Comparable projects listed for this tool include Danswer, AnythingLLM, Open WebUI and PrivateGPT. This guide does not compare their features or quality in detail.

How do I back up or upgrade DocsGPT?

Use docsgpt backup to back up the database and uploaded data, and docsgpt upgrade to upgrade and restart on the new version.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

DocsGPT suits teams and developers who want a private, self-hosted assistant over their own documents, and who can run Docker and manage upgrades and backups. If you only want a quick hosted chatbot with no setup, or you need guaranteed answer accuracy, pick a hosted service or another tool and test it first.

Install options & downloads