Installation

10xGraph needs Python 3.12 or newer. Install the framework and the API server with pip install 10xgraph 10xgraph-api, add an extra for your model provider, then check the install with 10xgraph version. The TypeScript client is a separate npm package.

What are the requirements?

How do I install the packages?

Use a virtual environment so the CLI and the libraries come from the same interpreter.

With pip:

python3.12 -m venv .venv
source .venv/bin/activate
pip install 10xgraph 10xgraph-api

With uv:

uv venv --python 3.12
source .venv/bin/activate
uv pip install 10xgraph 10xgraph-api

10xgraph is the core framework. 10xgraph-api adds the API server and the command-line tool.

Which extras should I add?

The base install has no model provider SDK. Add the extras you use, either in the same command or later.

pip install "10xgraph[openai]" 10xgraph-api

Combine extras with commas: pip install "10xgraph[openai,pg_checkpoint,mcp]". Other extras exist for Anthropic on Vertex AI and Bedrock (anthropic-vertex, anthropic-bedrock), SQLite checkpointing (sqlite_checkpoint), Qdrant and Mem0 memory stores, event publishers (kafka, rabbitmq, redis) and observability (otel, logfire, langsmith). The all extra installs most of them at once and is meant for development and CI.

Note: 10xGraph is pre-1.0. Pin the versions you deploy and read the changelog before upgrading.

How do I verify the install?

10xgraph version
python -c "import tenxgraph; print(tenxgraph.__file__)"

The first command prints the CLI version. The second confirms the library imports from your virtual environment. If the command is not found, activate the environment again and see the installation troubleshooting page.

How do I set provider API keys?

Providers read their keys from environment variables. Export them in your shell, or put them in a .env file that your 10xgraph.json points at with "env": ".env". Never commit that file.

export OPENAI_API_KEY="your-key"

How do I install the TypeScript client?

npm install 10xgraph-client

The client talks to a running 10xgraph api server. See Quickstart.

Scaffold a production project

10xgraph init --yes --template production --auth jwt --rate-limit redis

The production template creates your graph, a prompt-injection validator, evals, tests and an 10xgraph.json config with JWT auth and owner-only threads turned on. Use --auth custom instead to also get an auth/ module with a BaseAuth subclass to fill in. For a minimal start, skip this step and follow the Quickstart.

Serve it

10xgraph api

The server listens on http://localhost:8000 and exposes:

Something went wrong?

• Installation troubleshooting: pip failures, command not found, imports failing, ignored environment variables.
• API server troubleshooting: startup and runtime issues.
• Error codes: what each structured error means.

Next step

Build your first agent in the Quickstart.