Supported agents
Provider eligibility, model availability and limits are set by that provider. A chat subscription does not automatically include every coding agent or model. Workser does not pool subscriptions across providers.
Gemini CLI, Kimi Code, Grok Build and Muse Code are not currently offered in the supported-agent picker. Use an agent listed above.
Browse supported coding agents
Connect a third-party agent
1
Check installation
Open the agent connection flow. Workser checks supported CLI binaries and their authentication state. Having the provider’s desktop chat app installed is not the same as having its coding CLI.
2
Complete the provider's sign-in
Follow the installation or sign-in instructions shown for that agent. The embedded terminal uses the provider’s own login flow. Complete authentication yourself; do not paste provider credentials into a project conversation.
3
Choose an agent and model
Connect the agent, choose a model available to your account, and use a small first task to confirm the setup. Provider quota and authentication errors must be resolved with that account.
copilot and use /login inside its interface. Installation and sign-in checks are separate; an installed but unauthenticated CLI cannot run a task yet.
Use the built-in Workser Code agent
Select Workser Code in the agent picker. Workser supplies access to its supported models, so there is no separate provider login to complete. Paid model usage draws from your organization’s credit balance. Spark supports credit top-ups; Starter and Scale include monthly credits. There is no weekly free-agent allowance. See Plans and credits.The free path
The app download and new organization’s 7-day Spark cloud trial are free. Model usage is a separate choice:- An agent you already subscribe to uses that provider’s allowance.
- OpenCode can use a free model when one is available in your configured provider. Check the selected model and its limits before running.
- The built-in Workser Code agent uses credits for paid models, including during the cloud trial.
Choose models and reasoning effort
Use the model choices exposed for your connected agent. Default lets the CLI choose; a named model selects one explicitly. Where supported, you can also select reasoning effort. Availability depends on the agent, installed version and provider account.Configure your team
In Team, assign an agent and optional model to each role. Configure instructions, skills, tools, app connections and references for the work it should do. You can also choose a backup agent and model.agent run uses a configured role. agent spawn starts a temporary teammate using a supported connected agent. A backup helps when a primary agent is unavailable, but does not remove the backup provider’s own limits.
Roles share project state through tasks, documents, decisions and handovers. They do not share a single model context window. Include a clear objective and references when handing off work. See The AI team.