Use cases¶

A few worked examples of what people actually do with gdc. Each links to the guide that covers it in depth.

Interactive pair-programming¶

Work alongside the agent in the terminal app: ask questions about an unfamiliar codebase, then have it implement a change while you review and approve each sensitive step.

cd ~/my-project
gdc-tui

where is user authentication handled in this project?

…then:

add rate limiting to the login endpoint and update the tests

→ Interactive terminal

Plan before you change anything¶

For a bigger or riskier change, have the agent research and propose a plan first. It can't touch your code until you approve, so you review the approach before any edits happen.

gdc print --profile plan "plan how to migrate from the old config format"

→ Plan mode

Automated, unattended runs (CI & scripts)¶

Run a task from a script or CI job. The answer streams to standard output; add --auto-approve so the run can act without a human present (only in environments you trust).

gdc print --auto-approve \
  "run the linter and fix every warning it reports"

→ Headless CLI

Understand and audit past work¶

Because every run is saved, you can review exactly what the agent did — useful for code review, incident analysis, or just remembering what you tried yesterday.

gdc sessions list
gdc sessions show <id>

→ Sessions & projects

Connect your own tools and data¶

Give the agent access to your issue tracker, a database, or an internal documentation service by connecting an MCP server — the agent can then use those as tools during a task.

→ MCP servers

Large, coordinated tasks¶

For work that's naturally parallel — auditing many files, or a broad migration — the agent can split the job across several subagents and bring the results together.

→ Subagents & multi-agent

Run gdc as a server so several people or machines can connect to one instance, or so the terminal app can drive gdc running on another host.

→ Admin guide