AI Agents
Spawn Claude Code agents with full tool access. Every agent runs as a real CLI process.
Agent Launcher#
Open the Agent Launcher with Ctrl+Shift+A. From here you can select a built-in agent preset, create a custom agent with your own system prompt, choose a model (Opus, Sonnet, Haiku), and configure per-project MCP servers.
Built-in Agents#
DAEMON ships with six pre-configured agents:
| Agent | Purpose |
|---|---|
| Debug | Trace errors and fix stack traces |
| Security Audit | Scan for vulnerabilities and exploits |
| Code Review | Review code quality and patterns |
| Git | Commit, branch, and PR management |
| Test Runner | Write and run test suites |
| Solana | On-chain interactions and program development |
Each agent runs as an independent Claude Code process with full filesystem access, terminal capabilities, and tool use.
Custom Agents#
- Custom system prompts to define exactly how the agent should behave
- Model selection to choose between Opus, Sonnet, and Haiku based on your task
- MCP configuration to assign project-level or global MCP servers to each agent
MCP Management#
Toggle project-level and global MCP servers from the Claude panel or Settings. Changes are written to .claude/settings.json (project-level) and .mcp.json (global).
Agent Sessions#
Every agent session is tracked in the Sessions tab on the right panel. Session data includes start/end timestamps, commands executed, files modified, and token usage.
For Solana projects, sessions can optionally be recorded on-chain via the Session Registry.