Quickstart¶
Agentsflow is the production-grade control plane for coding agents. It provides a harness that controls workflow setup, routing, and guardrails so coding agents stay on rails. The recommended way to run AgentsFlow is as an MCP server using uvx.
Prerequisites¶
- uvx (from uv)
- Temporal CLI/dev server: install and run via https://docs.temporal.io/cli (the stack will auto-start a local dev server if port 7233 is free)
- Secrets:
OPENAI_API_KEYplus issue provider credentials (Jira or GitHub) - Coding agent CLI: install one of Claude Code, Codex CLI, or Gemini CLI (whatever assistant you plan to use)
1) Option 1: Use with AI Assistant (Recommended)¶
To use AgentsFlow with your AI assistant, configure it as an MCP server.
Claude Code¶
claude mcp add agentsflow --scope user --env OPENAI_API_KEY=sk-... --env PROCESS_CODING_AGENT_PROVIDER=claude -- uvx agentsflow
Gemini CLI¶
gemini mcp add agentsflow --scope user --env OPENAI_API_KEY=sk-... --env PROCESS_CODING_AGENT_PROVIDER=gemini -- uvx agentsflow
Codex CLI¶
codex mcp add agentsflow --env OPENAI_API_KEY=sk-... --env PROCESS_CODING_AGENT_PROVIDER=codex -- uvx agentsflow
Cursor¶
Add to .cursor/mcp.json or configure via Settings > MCP:
{
"mcpServers": {
"agentsflow": {
"command": "uvx",
"args": ["agentsflow"],
"env": {
"OPENAI_API_KEY": "sk-...",
"PROCESS_CODING_AGENT_PROVIDER": "codex"
}
}
}
}
Set PROCESS_CODING_AGENT_PROVIDER to pick which coding agent (claude / codex / gemini) AgentsFlow will delegate code actions to.
2) Option 2: Standalone HTTP Server¶
If you want to run the stack independently (e.g. for debugging or remote access via SSE/HTTP):
uvx agentsflow --transport streamable-http
3) Tracking Progress¶
Regardless of which option you choose, you can track workflow execution, history, and status in the Temporal Web UI.
- URL: http://localhost:8233
- Namespace:
default
3) Using AgentsFlow¶
Once connected, you can use natural language to interact with the process workflow. A few concrete prompts:
- "Call
start_agentsflow_processon this repo with issue https://jira.example.com/browse/TEAM-123." - "Kick off
start_agentsflow_processwith task text 'Fix 500 error when cookie is missing' for the current repo." - "The workflow paused; send these answers via
provide_agentsflow_clarificationand thenawait_agentsflow_resultwith the workflow id in a list (it returns after the first workflow finishes or pauses)." - "Check the latest status for workflow
process-1a2b3cusingawait_agentsflow_result([\"process-1a2b3c\"]); call again with remaining ids for multi-workflow checks."
You can supply either an issue URL or a free-text task when calling start_agentsflow_process, but not both.
Manual Setup (Alternative)¶
If you prefer to run from source or manage the process manually:
- Clone the repository.
- Install dependencies:
uv sync --extra dev. - Run the dev stack:
python -m agentsflow.dev.