> ## Documentation Index > Fetch the complete documentation index at: https://docs.flowfield.sh/llms.txt > Use this file to discover all available pages before exploring further. # Getting Started > Connect your coding agent, agree on a task, and review its result. Flowfield keeps your plan and each task's feed beside your coding conversation. Discuss work with your coordinator while workers implement agreed tasks in parallel. Use the board to see progress and the task feed to follow questions, answers and results. Flowfield board with tasks, priorities and work ready for review ## Install and start Install with [uv](https://docs.astral.sh/uv/getting-started/installation/): ```sh theme={null} uv tool install --python 3.12 flowfield-core flowfield serve ``` Open [localhost:8765](http://127.0.0.1:8765). Keep the terminal open; Ctrl-C stops the service. See [Installation](/installation) for the pip alternative. Managed workers also need Git and an installed, signed-in Codex CLI. ## Add your project In another terminal, change to an existing project's directory: ```sh theme={null} flowfield project init flowfield project guidance preview flowfield project guidance install flowfield integration connect codex ``` Registration adds `.flowfield/config.toml`. Guidance installation preserves existing instructions and adds a reference to `.agents/skills/flowfield-coordinator/SKILL.md`. Review the added files and commit them using your project's usual workflow before running workers. Flowfield adopts your project; it does not scaffold an application. Start a fresh Codex conversation in that directory. Ask: > Read this project's Flowfield board and give me its link. The agent should identify your project and return its board link. A new conversation loads the installed skill and MCP connection. See [Codex](/integrations/codex) for connection checks and troubleshooting. ## Agree on work Ask your coordinator to set up worker execution for the project: choose the target branch, dependency setup, checks and a local run command, then validate the environment. Choose a worker model and reasoning effort explicitly. These settings are also available under the settings icon beside the project name. Describe one small change and its success criteria. Ask the coordinator to capture and prepare the task in Flowfield. Open its feed to read the agreed definition. The coordinator maintains task intent; you control priorities and review. ## Run the task Move the task to **Up next** and click **Run queue**. Ready tasks run when their dependencies, setup and input requirements are satisfied. The task moves to **In progress**; its feed shows worker activity and any questions. Answer questions in the task feed or **Needs you**. Your saved answer lets the service continue managed work when the queue permits it. **Pause queue** prevents new starts; active workers continue. **Stop worker** stops the selected attempt. ## Review and finish When the task reaches **In review**, read its result and checks. Inspect the diff or use **Try result** to test an independent copy of that exact version. Record testing in the feed, request changes, or choose **Approve and integrate** for the selected result. A task result with its report, checks and review controls Flowfield delivers approved code to your configured project checkout, then marks the task **Done**. Local edits that prevent delivery remain intact and the result explains the next action. A complete report-only task can finish without code approval. To run independent tasks together, set **Maximum parallel workers** to **2** under **Workers**. Dependent tasks wait until prerequisite work is complete and its code is available. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.