# AI use cases for Retinue (jklthinking/retinue) Repo: [jklthinking/retinue](https://github.com/jklthinking/retinue) | Use cases: 3 | Page: https://dedupedai.com/use-cases/jklthinking-retinue/ Retinue is a self-hosted task board where humans and AI agents work the same cards, with holder-only writes, acceptance checks, and an append-only receipt chain. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Assign work cards to AI agents and check results on one board: https://dedupedai.com/use-cases/jklthinking-retinue/assign-cards-to-ai-agents-on-one-board/ - Connect a coding agent to the board through MCP: https://dedupedai.com/use-cases/jklthinking-retinue/let-coding-agents-use-the-board-through-mcp/ - See token activity and model quota per agent on the Home view: https://dedupedai.com/use-cases/jklthinking-retinue/monitor-agent-token-use-and-model-quota/ ## How to assign work cards to AI agents and check results on one board with Retinue Repo: [jklthinking/retinue](https://github.com/jklthinking/retinue) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Give each agent a card with one holder, an acceptance line, and a receipt trail you can review. **Who it is for:** A developer or team lead supervising several AI coding agents who needs one place to see who holds each task and whether it is really finished. **The problem:** Agent work gets scattered across chat threads and terminals, so nobody can tell which agent is doing what or whether a claimed result matches the request. **The outcome:** A running local hub where a card is claimed by one named agent, updated with progress and a result reference, and marked done only after its acceptance line is true. ### Steps 1. Export RETINUE_ADMIN_PASSWORD with at least 8 characters, then run `docker compose up --build` from the repo folder. 2. Run `curl -fsS http://127.0.0.1:9219/api/health` and confirm the JSON status is ok. 3. Sign in as operator, open Administration, and prepare an executor with actor id worker-1. Save its one-time token outside the data volume. 4. Open Task workspace, then Task board, then New task. Set the holder to worker-1 and write an acceptance line the agent can check. 5. Have the agent claim the card and post progress with its agent token as a bearer header to the card update endpoint, then set the status to done once the acceptance line is true. ### Prompt to paste into your agent ```text Using the Retinue repo at , start the hub with docker compose up --build and confirm /api/health returns status ok. Create a card for actor worker-1 with the acceptance line in , then claim it and post progress through the update endpoint. Report the task id and the final status, and verify that a token for a different actor gets 403 holder-only-writes. ``` ### Caveat The README's quick path stays on loopback, so reaching the hub from another machine needs the exposure steps in SELF_HOSTING.md and is not covered here. Risk, credentials: The one-time agent token and the admin password grant write access to the board, so store them outside the data volume and keep them out of cards. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/jklthinking-retinue/assign-cards-to-ai-agents-on-one-board/ ## How to connect a coding agent to the board through MCP with Retinue Repo: [jklthinking/retinue](https://github.com/jklthinking/retinue) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Let Claude Code or Codex read and update Retinue cards without hand-written curl calls. **Who it is for:** A developer who already runs Claude Code or OpenAI Codex and wants those agents to pick up and report on board cards directly. **The problem:** Agents need a way to claim work and write back that fits how they already operate, and hand-built HTTP calls are fragile to maintain. **The outcome:** An MCP-capable agent that can see its assigned cards, post progress, and hand off or complete work through the same holder-only rules as the HTTP API. ### Steps 1. Install the MCP extra with `pip install 'retinue[mcp]'`. 2. Make sure the hub is running (`curl -fsS http://127.0.0.1:9219/api/health` should return status ok). 3. Read docs/agent-onboarding.md in the repo to set up the agent identity and token. 4. Start the MCP server with `retinue-server mcp` and register it with the agent as an MCP server. 5. Ask the agent to claim one assigned card and post a progress update, then check the card's receipt chain on the board. ### Prompt to paste into your agent ```text In the Retinue repo at , install retinue[mcp], start retinue-server mcp, and register it with this agent. Claim the card and post one progress update. Report the claim and update receipts, and verify the card shows the agent as holder. ``` ### Caveat The README names the MCP command and onboarding doc but does not list the exact MCP client configuration, so the agent registration step depends on the agent's own MCP setup. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/jklthinking-retinue/let-coding-agents-use-the-board-through-mcp/ ## How to see token activity and model quota per agent on the Home view with Retinue Repo: [jklthinking/retinue](https://github.com/jklthinking/retinue) | Difficulty: Advanced | Time: 1 hour | Basis: From the README Compare how much each agent uses and how much subscription quota remains, with source labels. **Who it is for:** A person paying for several model subscriptions who wants to know which agents are consuming usage and when quotas reset. **The problem:** Usage is spread across Claude Code, Codex, and other subscriptions, so it is hard to tell which agent used what or when a limit will reset. **The outcome:** A Home view and Operations view that show reported token activity per agent and used percentages, reset times, and report age for enabled model quotas. ### Steps 1. Run `docker compose up --build` and confirm the hub with `curl -fsS http://127.0.0.1:9219/api/health`. 2. Upgrade an existing server through the explicit schema-26 migration described in docs/node-quota.en.md, since the README says existing servers require it. (inferred, check the README) 3. Enable the node quota collector locally for the subscriptions you use, since collectors stay off until enabled. (inferred, check the README) 4. Open the Home view and check the Claude, Codex, Grok, Cursor, or Kimi cards for used percentage, reset time, and report age. 5. Open the Operations view and check reported usage coverage and receipt freshness before trusting the per-agent totals. ### Prompt to paste into your agent ```text In the Retinue repo at , enable the node quota collector as described in docs/node-quota.en.md and open the Home view. Report the used percentage, reset time, and report age for each enabled subscription, and verify that the Operations view labels each source and its coverage. ``` ### Caveat The README says reported usage is not a complete provider bill, so these numbers should be read as coverage-limited estimates rather than invoices. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/jklthinking-retinue/monitor-agent-token-use-and-model-quota/