Skip to main content

AI Agents

Gitghost AI supports AI workflows while keeping project policy, review, and evidence in the loop. Use this page as the overview, then follow the detailed guides for chat, local agent pairing, sessions, approvals, and guardrails.

Gitghost AI Agents page

Modes​

ModeWhere Work HappensPrimary Use
Hosted Gitghost AIGitghost AI chat and managed agent sessionsPlanning, review, documentation, bug investigation, and proposed changes.
Local agent syncYour local coding tool plus Gitghost AI evidence syncContinue using tools such as Codex, Claude Code, Gemini CLI, Cursor, OpenCode, Copilot CLI, or custom adapters while Gitghost AI records review evidence.

Built-In Agent Roles​

Depending on your workspace, Gitghost AI may show roles such as:

  • Code Assistant for general implementation help.
  • Code Reviewer for bugs, security issues, and maintainability.
  • Bug Fixer for diagnosis and small fixes.
  • Documentation Writer for project documentation.

Start with narrow prompts. Include the goal, affected area, constraints, and expected verification.

Local Agent Flow​

  1. Install and authenticate the CLI.
  2. Open the project in Gitghost AI.
  3. Start or select an agent session.
  4. Copy the connect code shown in the UI.
  5. Run the connect command shown by Gitghost AI.
  6. Continue working locally.
  7. Review session evidence, checkpoints, action requests, and merge requests in Gitghost AI.

See AI chat and agent pairing.

Review Agent Work​

Review AI work like a teammate's pull request:

  • Read the original task.
  • Inspect changed files.
  • Check linked issue and branch.
  • Confirm tests, pipelines, and security checks.
  • Review transcript or evidence when available.
  • Approve action requests only when they match the scope.

Policy Controls​

Project settings can control:

  • Whether local agent sync is allowed.
  • Which agent types are available.
  • Whether mutating actions need approval.
  • Whether command-capable tools are allowed.
  • Whether tools that may access sensitive project data are allowed.
  • Budget and retention limits.

Known user-facing policy block:

Local agent sync is disabled by project Agent Policy

If you see that message, ask a maintainer to review Project Settings > AI Settings.

Detailed AI Guides​