Guide · Agent skill

How to use the cognee-docker skill to run cognee in Docker

Updated 2026-10-29 · written with AI from the project's own documentation and checked against our catalogue data

cognee-docker is an agent skill for the cognee project, a tool that turns documents into AI memory. It teaches your AI agent how to start cognee inside Docker containers, from a single prebuilt image to the full repository stack. This guide explains what the skill does, how to install it, how to ask for it in plain language, and what to check before you run it.

Run cognee in Dockertopoteretes/cognee · Apache-2.0 Download the skill

Quick start

1Download and unzip the skill. Download the cognee-docker ZIP from this page and unzip it so the folder contains SKILL.md.
2Install it for your agent. Place the folder in ~/.claude/skills/ or a project's .claude/skills/, or upload the ZIP under Settings, Capabilities, Skills in the Claude apps.
3Make sure Docker and an OpenAI key are ready. Confirm docker compose works on your machine and that you have an OpenAI API key to use as LLM_API_KEY.
4Ask the agent for a try-out setup. Request a local cognee run in Docker without cloning or building. The agent writes a single docker-compose.yml that pulls cognee/cognee:main.
5Start the container and check health. Export LLM_API_KEY, run docker compose up, then request http://localhost:8000/health to confirm the API is running.
6Send a first note and query. Use the remember and recall endpoints to ingest a short file and ask a question. Browse http://localhost:8000/docs for the full API reference.

What the cognee-docker skill makes the agent do

When the skill is installed, the agent knows the commands and settings needed to run cognee with Docker or docker compose. Its description covers three jobs: trying cognee from the prebuilt image, starting the API server in a container, and bringing up the full stack with UI, MCP server, Postgres and Neo4j through compose profiles.

The skill is a setup guide. It tells the agent which files, commands and settings to use. It does not add new analysis of its own.

Write a minimal docker-compose file that pulls the prebuilt cognee/cognee:main image, with no clone or build step
Walk you through starting the API server, checking its health endpoint and sending a first note and query
Set up the repository's docker-compose.yml with optional profiles for the UI, MCP server, Postgres and Neo4j

How to install the cognee-docker skill

You need the skill folder on your machine. Download the cognee-docker ZIP from the download link on this page, then unzip it. The folder contains SKILL.md, the file that describes the skill to the agent.

For Claude Code, place the unzipped cognee-docker folder in your personal skills directory at ~/.claude/skills/, or in a project's .claude/skills/ directory if you want the skill only for that project.

In the Claude apps, you can instead upload the ZIP under Settings, then Capabilities, then Skills.

Personal install: ~/.claude/skills/cognee-docker/
Project install: .claude/skills/cognee-docker/ inside your project folder
Claude apps: Settings → Capabilities → Skills, then upload the ZIP

How to ask the agent to use the skill

You do not need to name the skill. The skill's description tells the agent when to use it, so plain requests about running cognee in Docker should match. Say what you want to achieve: a quick try-out, a single API server, or the full stack with extra services.

If the agent does not pick up the skill, mention cognee and Docker explicitly in your request.

Set up cognee in Docker so I can try it locally, I don't want to build anything.
Run the cognee API server in a container and show me how to send a first note.
Bring up the full cognee stack with docker compose, including the UI and Postgres.
Start cognee with the Neo4j profile and tell me the login details it uses.

Run the prebuilt image with one compose file

For a local try-out, the skill does not clone or build anything. It has the agent save a small docker-compose.yml in an empty directory. The file pulls cognee/cognee:main and maps port 8000. It requires LLM_API_KEY and turns off backend access control for a single-user try-out.

After saving the file, the agent sets your OpenAI key, starts the container and checks the health endpoint. The interactive API reference is at http://localhost:8000/docs. The first requests store a small text file and ask a question about it.

remember ingests a file and builds the graph in one call
recall queries your data; omit search_type to let cognee route the query automatically
Request fields accept both snake_case and camelCase, so search_type and searchType both work
Data stays inside the container unless you set DATA_ROOT_DIRECTORY and SYSTEM_ROOT_DIRECTORY and mount a named volume at /cognee-data

export LLM_API_KEY="sk-..."
docker compose up
curl http://localhost:8000/health
echo "Cognee turns documents into AI memory." > note.txt
curl -X POST http://localhost:8000/api/v1/remember -F "data=@note.txt" -F "datasetName=main_dataset"
curl -X POST http://localhost:8000/api/v1/recall -H "Content-Type: application/json" -d '{"query": "What does Cognee do?", "datasets": ["main_dataset"]}'

Run the full stack from the repository

The repository's own docker-compose.yml builds from source and adds opt-in profiles. From the repo root, the agent needs a .env file with at least LLM_API_KEY, which it creates by copying .env.template. Without profiles, only the API server on port 8000 starts.

Profiles add services. The UI runs on port 3000 and the MCP server on port 8001. The Postgres profile uses pgvector/pg17 with user, password and database cognee on port 5432. The Neo4j profile uses user neo4j with password pleaseletmein on ports 7474 and 7687. If cognee runs in a container and the database runs on your host machine, the skill sets DB_HOST to host.docker.internal.

docker compose up
docker compose --profile ui up
docker compose --profile mcp up
docker compose --profile postgres --profile neo4j up

Safety notes: access control, credentials and network use

The try-out configuration is for a single user on your own machine. It sets ENABLE_BACKEND_ACCESS_CONTROL to false, which means the API does not require authentication. It also uses shared local databases with default credentials. Do not use these settings with real or sensitive data, and do not expose the ports to a network you do not control.

The skill does not run scripts on your computer. It does use the network: Docker pulls the cognee/cognee:main image, and cognee calls your LLM provider with your API key. Your documents are sent to the provider as part of processing, so review what you upload.

Access control is on by default, so every API call needs authentication unless you turn it off
Default Postgres and Neo4j logins are public in the skill and should be changed before any shared use
Keep LLM_API_KEY out of files you share or commit to version control

What people use it for

Trying cognee on a laptop to see how it turns a text file into a queryable memory, without cloning the repository
Starting a single cognee API server in a container for a developer to test integrations against
Bringing up the repository stack with the UI and MCP server for a small team to share
Running cognee with Postgres and Neo4j profiles to test how it works with those databases

Frequently asked questions

How do I install cognee in Docker with this skill?

Download the cognee-docker ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. Then ask your agent to run cognee in Docker, and it will create a docker-compose.yml that pulls the cognee/cognee:main image. You need Docker, docker compose and an OpenAI API key.

What do I need to run cognee with Docker?

You need Docker with docker compose, an OpenAI API key set as LLM_API_KEY, and internet access to pull the image and call the LLM provider. The default configuration uses OpenAI for both the language model and embeddings, so a key from another provider alone is not enough unless both are configured.

Is cognee-docker free to use?

The cognee project is licensed under Apache-2.0, and this skill is part of its repository. The skill source does not cover pricing, but calls to the OpenAI API made with your key are billed by OpenAI, so check their pricing before you run it.

Why does my cognee API request return an authentication error?

By default, backend access control is on and every API call requires authentication. The single-user try-out sets ENABLE_BACKEND_ACCESS_CONTROL to false to skip this. Use that setting only for local testing with non-sensitive data.

Can I use cognee with Docker in production with this skill?

No. The skill's try-out configuration disables access control and uses shared local databases with default credentials, so it is not suitable as-is for real data. Use it to learn the setup, then change the credentials and security settings before any shared or production use.

What is the difference between this skill and running cognee directly?

This skill is only a setup guide for Docker. It tells the agent which commands, files and settings to use for the prebuilt image or the repository compose stack. It does not replace the cognee project's own documentation for its features.

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

Use cognee-docker if you want to try cognee on your own machine or set up a small shared server with Docker Compose, and you already have Docker and an OpenAI key. Pick something else if you need a production deployment with real data, or if you do not want to use Docker or an LLM API key.

Download the skill