ruClip is a ruvnet reinterpretation of the ideas behind paperclipai/paperclip: a control plane that sits above whatever agents you already have — org chart, goals, budget-gated heartbeats, task/approval workflow, and a signed audit trail — rather than an agent framework in its own right. paperclip does not itself automate sales/eng/support/finance; neither does ruClip. What it provides is the governance substrate for running agents as if they were employees of a company with a goal, a budget, and accountability.
ruClip is built entirely on tools ruvnet already owns and operates. Nothing here reimplements what already exists elsewhere in the ecosystem:
| Capability | Backed by |
|---|---|
| Agent orchestration, work-ownership, signed audit trail | ruflo — Agent Teams, claims_*, witness |
| Build-time governance of ruClip's own codebase | metaharness — metaharness score/genome, a custom ruclip-metaharness .harness/bench.json suite. @metaharness/redblue/@metaharness/flywheel are explicitly not part of this — that role belongs to Autogenous (below), permanently |
| Runtime governance of the live company | ruvnet/autogenous — antibody/canary/rollback model, packages/radio-moe as the cross-provider agent adapter |
| Semantic memory + Company/Goals/Issues schema | ruvector + AgentDB |
| Human+agent comms (the Slack-equivalent) | agentbbs |
| Nightly evidence-gated evolution | dream-machine — config only, no new scheduler |
Edge/offline agent transport (LatentMesh)
is intentionally not on the v1 critical path — it solves connectivity, a
different problem than company orchestration. See
ADR-0001 for the full reasoning
and docs/PLAN.md for the phased implementation roadmap.
Active development. Shipped: Phase 1 (control-plane core — Company/Goals/
Issues schema, approval-gate state machine, claims_*-backed
authorization, ActorCredential identity verification, budget-gated
heartbeats, agentbbs comms with radio-moe signing, per-OrgMember
interaction profiles), Phase 3 (build-time governance genome), Phase 2a
(read-only company-board dashboard, as a Claude Artifact), and Phase 2b's
first slice — a real HumanIdentityAttestation → ActorCredential
minting primitive, so a human OrgMember holding a valid attestation can
already call applyApprovalTransition end to end. Still open: no
attestation producer exists yet (no login/dashboard flow), which is now
Phase 2b's narrowed remaining scope, and dashboard write actions
(approve/reject UI) aren't built. See docs/PLAN.md §8 for the full phased
roadmap and current delivery status, and docs/design/ for every slice's
design doc.
MIT — see LICENSE.