OverCR
GuideboardLabs/overcr
Hermes-first orchestration substrate with subagent routing, approval gates and sandboxed execution
OverCR is a Hermes-first operational substrate that AI workloads run inside, rather than a chatbot or product of its own. It provides identity, subagent routing, filesystem-first state, approval gates and audit trails for agent work.
What OverCR does
OverCR describes its scope as identity, doctrine, boot continuity, subagent routing, packet validation across six levels, workflow choreography, audit trails, approval gates and semantic memory. Its capabilities were built up in versioned layers: a substrate core with task lifecycle and approval gates, a governed semantic memory layer, a terminal operator TUI that reads from the filesystem and never mutates it, a reusable workflow library, a knowledge layer with source registry and provenance tracking, and operator-auditable single-page web ingestion with prompt-injection scanning.
Later versions add an execution sandbox with a 14-command allowlist, filesystem and network guards and rollback snapshots, kernel isolation backends such as Bubblewrap and Firejail chosen automatically, composite workflow DAGs with branching and retry policies, and integration-hardening validators. The newest layer reads structured facts from an Obsidian vault without a vector database or embeddings, and tracks rejected approaches and next actions as facts so agents stop retrying dead ends.
Key features
- Subagent routing with multi-level packet validation and approval gates
- Filesystem-first semantic memory with rule-gated promotion and conflict detection
- Terminal operator TUI that reads state without mutating it
- Execution sandbox with a command allowlist, network guard and rollback snapshots
- Automatic kernel isolation backend selection (Bubblewrap, Firejail)
- Obsidian vault integration for context, with negative-fact and next-action tracking
When to use it
- Running Hermes-based agent work under an auditable, approval-gated execution layer
- Giving agents governed read access to an Obsidian vault of project knowledge
- Tracking dead-end approaches as negative facts so agents do not retry them
Who it is for: Teams who want a governed, auditable operating layer under Hermes Agent rather than an ungated agent loop.
How it fits with Hermes Agent
Described in the README as a Hermes-first portable AI orchestration substrate that Hermes-based agent workloads run inside.
FAQ
What is OverCR?
OverCR is a Hermes-first operational substrate that AI agent workloads run inside. It adds identity, subagent routing, approval gates and audit trails rather than acting as a chatbot itself.
Does OverCR work with Hermes Agent?
Yes. The README describes OverCR as Hermes-first, meaning it is built around Hermes-based agent workloads as its governed operating layer.
Is OverCR free and open source?
Yes. The repository is released under the MIT license.
Similar multi-agent for Hermes Agent
All multi-agentAgent registry implementing ARD v0.9, operated by a Hermes Agent, with free and hosted listings
asimons81 Hermes A2A BridgeArchived plugin that let Hermes Agent call and act as A2A peers, now replaced by built-in A2A
xuyang-liu16 Hermes Code BridgeHermes Agent plugin that routes work to local coding agents like Codex, Claude Code, Kimi Code and OpenCode
amajorai RyuComposable layer for building and running AI agents, with a gateway, CLI, desktop app and plugins
SouthpawIN Agent WikiMarkdown-vault multi-agent system built on Hermes Agent, with skills, agents and loops per topic
keeltrace NerveSupervisory layer for Hermes Agent with typed decisions, verified task completion and an opt-in tool gate
Related guides: Build agent teams with the Hermes kanban board · Run multiple Hermes agents with profiles · Connect Hermes agents on several machines with Hermes Desktop