Modus Core

Integrations

Integrations for open agent runtimes.

Modus Core plugs into existing open agent stacks and turns their activity into measurable traces, workstate memory, verifier-backed evaluations, and improvement loops.

Keep your runtimeUse OpenClaw, Hermes Agent, Claude Code, Codex, Goose, Cline, or your own stack.
Add the missing layerEvaluation, memory, verifier rewards, governance, and human control.
Export the signalTurn sessions into traces, failure taxonomies, evals, and RL-ready trajectories.

Runtime compatibility

We instrument the tools teams already use.

Modus Core is designed to integrate with open agent runtimes through skills, hooks, logs, gateway events, shell wrappers, MCP tools, and trajectory exports. It is not a demand to replace the runtime.

OpenClawHermes AgentClaude CodeCodexGooseClineCursor and Windsurf-style coding agentsCustom LangGraph, MCP, and shell-based agents

Runtime-specific paths

OpenClaw and Hermes Agent are natural first targets.

OpenClaw

Observe broad self-hosted agent deployments

For OpenClaw deployments, Modus Sentinel can observe gateway and session activity, tool usage, skills, cron jobs, and channel-triggered workflows. Modus Memory preserves the surrounding local workstate, while Modus Verify attaches external tests and verifiers to evaluate outcomes.

Hermes Agent

Turn research-native agent activity into measurable trajectories

For Hermes Agent deployments, Modus Core can sit around the gateway, terminal backends, skills, memory files, RPC tools, and batch trajectory paths. The goal is to convert self-improving agent activity into traces, failure taxonomies, verifier-backed rewards, and reusable training data.

Integration surfaces

Where Modus can connect.

Skills and tools

Installable skills, tool calls, command wrappers, and runtime extensions that expose what the agent tried to do.

Hooks and logs

Session hooks, terminal logs, repository events, test output, file diffs, retries, and failure records.

Gateway events

Messages and commands flowing through Discord, Slack, Telegram, email, CLI, and other agent-facing channels.

MCP and API boundaries

Model Context Protocol tools, HTTP APIs, RPC paths, and local services that can emit structured traces.

Terminal backends

Local shells, Docker, SSH, Singularity, Modal, cloud sandboxes, and other execution environments.

Trajectory exports

Prompt, action, observation, artifact, score, and verifier records suitable for evals, analysis, and RL datasets.

Research use cases

Useful for teams studying real agent behavior.

  • Agent safety evaluations across long-running local tasks
  • Coding-agent evals grounded in files, diffs, tests, and terminal output
  • RL trajectory generation from real agent attempts
  • Verifier-guided improvement using tests, Lean, typecheckers, and domain validators
  • Long-horizon task analysis with failure taxonomies and recovery traces
  • Human supervision through Pairling and Modus Workbench

CTA

Bring us your existing agent stack.

If your team already uses open agent runtimes for research, automation, coding, or safety work, Modus Core can help make that activity observable, comparable, and improvable.