Matrix recipes
Use Matrix recipes as reusable task briefs for research, reports, coding, and everyday work, with clear inputs, outputs, and review boundaries.
Maintained by the Matrix OS team. Product instructions are updated with the supported workflow.
A Matrix recipe is a reusable task brief. It describes the inputs an agent needs, the work to perform, the expected result, and what you should review.
Browse Matrix recipes for research, weekly reports, bug fixes, campaign briefs, meeting follow-ups, and software spend reviews.
Use a recipe today
- Open a recipe and choose Open in Matrix. Public inspiration cards use Build in Matrix.
- On a computer with recipe handoff support, Matrix opens a new Chat draft for the selected recipe. Review the draft and choose your model before sending.
- You can also choose Copy recipe and paste the instructions into Chat or Electron Desktop. If copying is blocked, expand Read the full instructions and copy the visible text.
- Provide the requested files or source material. Connect required tools through their supported authentication flows.
- Review the result and its sources before using it or approving an external action.
These are manually started task briefs. They do not install a bot, connect an account, or create a schedule. Results depend on the agent, tools, and inputs you provide. Missing access should be reported as a blocker.
Opening a recipe does not send the draft or install an Agent. If an older Matrix computer opens without the draft, use Copy recipe instead.
On computers where Chat Agents is enabled, you can ask Chat to create a reusable Agent. Agree on its role, then review the saved Agent through Manage agents in the Chat rail. Available Codex and Hermes models are supported. A saved Agent appears in the rail; selecting it prepares a request with its configuration. Confirm its access before sending. If setup or saving fails, Chat should report the blocker rather than claim the Agent is ready.
Jev Inbox Triage Bot
On computers running the updated Jev recipe, choose Jev Inbox Triage in Matrix's recipe catalog and use Set up bot. This creates a Matrix Pi Bot with its own conversation. It does not require Hermes or a personal Jev API key. Jev classifications use Matrix AI credits; the Bot's main model follows its supported Matrix model configuration.
Choose your connected Gmail account when the Bot asks. Review the requested access before selecting it: read the Inbox and add Jev classification labels. This permission preserves existing labels and does not allow archiving, sending, deleting or marking messages read. Connecting Gmail alone does not grant this Bot access, and permissions from older Agents are not copied automatically.
You can ask: “Classify my entire Inbox by email type and add the matching Jev labels. Preserve existing labels and report confirmed results.” The Bot processes the Inbox in resumable batches. If a run stops before finishing, ask it to resume the saved batch. A progress report should distinguish confirmed labels, unchanged threads, items needing review and unconfirmed actions; unconfirmed does not mean successful.
An older read-only Inbox Bot remains read-only. Create the updated recipe Bot to use verified labeling, and revoke its Gmail access from the Bot's access panel when you no longer want it to continue.
Repeat a successful task
Start with a single run. Refine the instructions after reviewing the output, then save your preferred brief in your workspace. Scheduling is a separate setup and requires an eligible plan and runtime capability. Always-on computers start with Starter. Scheduling remains a separate capability; verify the selected plan and runtime support before relying on a recurring workflow.
Recipes, agents, skills, and schedules
- A recipe describes a job and the result you want.
- An agent does the work using the tools available to it.
- A skill supplies reusable instructions for a capability.
- A schedule determines when a recurring task runs.
A recipe can reference skills and be used for recurring work, but copying its text does not grant permissions or enforce a security policy. Review the chosen agent's permissions and approve external actions deliberately.
Keep credentials out of recipe text. Share reusable instructions without private source files, account tokens, or previous run results.
How is this guide?