Capy & Multi-Agent Grids

Capy is Sandwalk's built-in meta-agent (⌃⌥C or the Capy button at the bottom-left of the sidebar). Unlike a regular session agent that works inside a single chat, Capy has full awareness of your active sessions, projects, Specialists, Memory/Wiki/Omnigraph, external coding CLIs, and benchmarking harness.

Parallel Multi-Agent Split Grids

Ask Capy to split a complex question across multiple sessions or Specialist personas at once:

• "Spawn a 2-session split grid with Nova and Soren to debate oral GLP-1R titration vs. hERG safety."
• "Create 4 parallel chats investigating KRAS G12D resistance mechanisms across different models."
• "Summarize session #abc123 and seed a new chat with Otium from its findings."

What Capy Can Do

• Session & Grid Orchestration: capy.spawn_sessions (with specialists="Nova,Soren"), capy.open_split, capy.seed_session, capy.read_session, capy.summarize_session.
• Memory, Wiki & Skill Stewardship: capy.memory_manage, capy.memory_search, capy.wiki_digest, capy.distill_session, and capy.create_skill_from_memory.
• External Coding Agents: Import sessions from or hand off chats to Claude Code, Codex, OpenCode, fx, and Hermes.
• Benchmarking & Diagnostics: Launch sandwalk eval suites (capy.bench_launch), inspect results (capy.bench_inspect), or run capy.doctor.