01
RepoOps
A governed repository-operation domain: structured requests, durable jobs, bounded implementation candidates, controlled adoption and verification.
Explore RepoOps →BARANES
Governed infrastructure for AI agents that act on real systems.
AI agents can plan, call tools and produce candidate changes. Baranes is designed to keep the path from intent to accepted effect bounded, traceable and reviewable.
RepoOps is the first concrete Baranes domain used to explain the model: repository work moves from bounded candidate execution to controlled canonical adoption.
The problem
Autonomy is easy to demonstrate. Controlled effects are harder.
AI agents can reason, call tools and modify code. The difficult boundary begins when an action must be attributable, bounded, reviewable and controlled before it becomes a real system effect.
“The agent produced a result” is not the same as “the system may accept the effect.”
Baranes is organized around that separation: intent, policy, execution capability, candidate state, evidence and canonical adoption remain distinct.
Bounded executionScope, paths, capabilities and stop conditions are explicit.
Effect governanceCandidate work is separated from canonical system effects.
Evidence firstExecution results, verification and provenance remain traceable.
Architecture
One system, several boundaries.
Baranes is broader than repository operations. RepoOps is the clearest first surface for showing the model end to end.
02
SDD Framework
The open-source lifecycle and contract layer used to make feature state, gates and regression paths explicit and machine-readable.
Open-source foundation →03
Baranes Atlas
Technical depth for people who want to inspect the target system, its modules, flows and authority boundaries.
Open technical Atlas →Governed flow
Keep cognition flexible. Govern the effect boundary.
IntentWhat is being asked and why
SpecificationExpected behaviour and acceptance
CandidateBounded engineering work
AdoptionControlled canonical effect
EvidenceVerify, audit, retain provenance
Current evidence
Start with what can be inspected.
RepoOps Flow Atlas
An interactive target model showing the SDD lifecycle, persistent artifacts, RepoOps job engine, role-level execution surface, Git adoption boundary, governance primitives and identity trace.
Open source
SDD Framework
Fresh-install SDD workflow layer with machine-readable schema and protocol, validator, installer, tests and CI.
View on GitHub ↗