A backend developer agent handles server-side implementation work: writing business logic, designing service boundaries, and maintaining the data layer. In a fleet, it picks up tickets marked ready and produces pull requests without waiting for a human to start the work.
A role-specific prompt matters because backend work has strong conventions: how your team structures service layers, which error types get wrapped vs. surfaced, whether you use repository pattern or active record. A generic coding agent ignores those contracts; a backend agent tuned to your codebase follows them from the first commit.
What this agent owns
- Implement API endpoints, service methods, and background jobs from ticket specifications
- Write and maintain unit and integration tests for business logic
- Review data models and migration files for correctness before opening a pull request
- Enforce service-layer contracts and error-handling conventions across changes
- Respond to code review feedback and iterate on pull requests until merged
Recommended model: Claude Sonnet
Most backend implementation tasks are well-defined enough that Sonnet produces correct, idiomatic code without needing Opus-level reasoning.
Example tasks
- Add a paginated list endpoint with cursor-based pagination
- Refactor a God object into smaller services with clear interfaces
- Write a database migration and accompanying rollback script
- Implement a retry queue for failed webhook deliveries
# create an agent from this template, then start it
$ fleet agent create --name backend-developer--vendor claude-code --template <template-name>
$ fleet agent start backend-developer
Find the exact template name with fleet template list.
