The capability model#
Not every agent exposes the same surface, and AgentPeek never pretends otherwise. Each integration is described along four capabilities:
- Monitoring: live sessions with state, activity, and timelines. Every supported agent has this.
- In-notch answers: permission prompts, questions, and plans answered from the notch. This works for Claude Code, Codex, Qoder, CodeBuddy, Gajae Code Direct Chat, OpenCode, MiMo Code, GitHub Copilot, Kilo Code, Factory Droid, Antigravity, Qwen Code, Pi, Oh My Pi Direct Chat, Cline, Goose Direct Chat, Kiro Direct Chat, Mistral Vibe Direct Chat, and Devin.
- Follow-up prompts: continuing a resumable session from the composer. This works for Claude Code, Codex, Cursor, Qoder, Gajae Code, Grok Build, Kimi Code, Hermes, OpenCode, MiMo Code, GitHub Copilot, Kilo Code, Factory Droid, Antigravity, Qwen Code, Pi, Oh My Pi, Goose, Kiro, Mistral Vibe, Devin, and Command Code.
- Usage: token, cost, or limit readouts. Usage readouts cover Claude Code, Codex, Cursor, GitHub Copilot, Kimi Code, Hermes, OpenCode, MiMo Code, Kilo Code, Antigravity, Qwen Code, Pi, Oh My Pi, Goose, Kiro, Cline, Mistral Vibe, ZCode, Devin, Command Code, and Aider when its analytics log is already configured.
Cursor, Grok Build, Kimi Code, Hermes, Oh My Pi, and ZCode surface prompts as observation-only attention flags. Approvals stay in the native agent. When an agent's records expose no usage, its sessions simply appear without usage numbers; AgentPeek never invents a quota, reset window, or cost.
Per-agent guides#
Every agent has its own page covering how it connects, what its sessions show, its prompt boundaries, and its usage sources:
Qoder, CodeBuddy, and Gajae Code#
Qoder hooks report activity and answer supported mutating-tool approvals, with selected-session ACP for Direct Chat. CodeBuddy hooks can return permission decisions. Gajae Code uses SDK discovery with process fallback, then resumes the selected session through ACP.