Works with the agents and runtimes you already use.
sys9 doesn't ask you to switch agent loops. Keep your runtime — sys9 brings the memory, sandbox, data, identity, and observability around it. Atomic services, opt in where you need them.
The ecosystem sys9 plugs into.
Nine runtimes, one substrate. Each integration is built to keep your agent loop intact.
Claude Code
First-class across the stack: mem9 memory, owl9 capture with one env var, and smith9 hosts it out of the box.
OpenAI Codex
mem9 memory and owl9 session capture support Codex; smith9 runs Codex agents out of the box.
OpenClaw
Plugs into mem9 for persistent cross-session memory that outlives resets and restarts.
Cursor
Bring your editor loop — compose sys9 sandboxes, data, and observability underneath it.
OpenCode
mem9 memory integration, and smith9 hosts OpenCode agents out of the box.
Cline
Bring your agent loop — sys9 provides the compute, data, and identity substrate around it.
Dify
Connects to mem9 so pipelines and apps share one memory across agents.
Hermes
Integrates with mem9 for shared, inspectable memory across sessions.
Pi
chord9 orchestrates Pi alongside Claude Code and Codex into one heterogeneous agent team.
Extend agent observability with community plugins.
owl9 captures agent sessions with zero config and ships a plugin marketplace
(owl9-plugins) so you can extend how runs are traced, transformed, and
surfaced. Build a plugin for your own workflow, or pull one the community already shipped.
# wrap the binary — no agent changes OWL9_ENABLE=1 claude ✓ session captured ✓ sub-agents auto-traced ✓ transcript ready (human-readable)
Building an agent or runtime? Let's integrate.
If your tool runs agents, sys9 can give it memory, compute, data, and observability. We'd like to hear from you.