Skip to content

Quickstart⚓︎

Prerequisites

  • Python 3.10+
  • uv package manager
  • API keys for at least one LLM provider (OpenAI, Anthropic, Google, etc.)

1. Install ContextAgent⚓︎

git clone https://github.com/context-machine-lab/contextagent.git
cd contextagent
uv sync

pip install contextagent

2. Configure providers⚓︎

Copy the template .env and add your keys.

cp .env.example .env

Update the following fields at a minimum:

Variable Description
OPENAI_API_KEY Required for OpenAI models
ANTHROPIC_API_KEY Needed for Claude agents
GOOGLE_API_KEY Enables Gemini support

Warning

Do not commit .env. Keep your provider credentials private.

3. Run an example pipeline⚓︎

uv run python -m examples.web_researcher \
  --topic "Latest transformer architecture techniques"

The pipeline spins up researcher + writer agents, fetches web context, and delivers a concise report.

4. Explore the pipeline manager UI⚓︎

uv run python frontend/app.py --host localhost --port 9090 --debug

Visit http://localhost:9090 to trigger pipelines, stream logs, and monitor artifacts.

5. Build your own agent⚓︎

Create a class that inherits from BasePipeline and implements run.

from pipelines.base import BasePipeline

class SupportPipeline(BasePipeline):
    def __init__(self, config_path: str):
        super().__init__(config_path)

    async def run(self, query):
        ticket = await self.context.agents["triage"](query)
        summary = await self.context.agents["writer"](ticket)
        return summary

Next, point the pipeline at a YAML config that defines the agents, tools, and prompts you need. Continue in the Custom Tool guide.

Where to go next⚓︎