Docs

AI Agents

A team of specialized agents that read, write, and run code.

The agents

Swork Team ships a set of role-specific agents that can work together on your codebase. Each has its own tools, system prompt, and permissions.

Team Lead — coordinates work across agents.
Product Manager — turns goals into tasks.
Architect — designs structure and interfaces.
Engineer — implements and edits code.
Tester — writes and runs tests.
Code Reviewer — reviews diffs for issues.

These six are a starting point — you can also design your own custom agents with their own prompts, tools, and models, covered below.

Models

Agents run on Claude, GPT, and Gemini, plus local models. Choose a model per agent, or set a default.

On a free plan you bring your own provider API keys. Paid plans include managed credits so you do not need keys.

Permissions

Each agent has configurable permissions for reading files, writing files, and running commands. By default, file writes and command execution surface a diff or confirmation before anything is applied.

Build your own agents

The six roles above are a starting roster, not a limit. You can create your own agents — give each a name and role, its own system prompt, a default model, the tools it may use, and read/write/run permissions.

Custom agents work exactly like the built-ins: they pick up tasks from the project board, hand work off to the other agents, and surface their changes as reviewable diffs. Spin up a Security auditor, a Docs writer, a DevOps engineer — whatever your projects need.

Creating custom agents is available on every plan, including the free Solo plan. Ultra adds shareable custom-agent templates for your team.