Overview
Claim Plane is a coordination architecture designed to manage concurrent software development by parallel coding agents. It addresses potential interferences that can arise when agents independently produce locally valid changes, such as expanding beyond planned scope, interfering at integration time, or relying on invalidated premises. The architecture models concurrent software change as a pre-write admission problem.
Research Context
Existing approaches to coordinating parallel coding agents typically focus on communication, isolated workspaces, late merge-time repair, continuous supervision, or post-hoc runtime recovery. These methods aim to mitigate issues where independently developed changes can conflict or invalidate each other during integration. Claim Plane introduces a pre-implementation admission mechanism to address these coordination challenges.
Approach
Claim Plane operates by requiring each worker to declare a versioned ChangeIntent prior to implementation. This ChangeIntent specifies an exact base commit, typed resources, dependencies, and operations. Operations within an intent are marked as either committed or contingent. A deterministic control plane is responsible for atomically admitting compatible intents. This control plane restricts same-file parallelism to regions declared within the intents. It also serializes unresolved overlaps, tracks dependency invalidation, and is designed to fail closed in cases of ambiguous authority.
A key aspect of Claim Plane's execution model is its handling of contingent mutations. Initially, a contingent mutation does not reserve write ownership. Instead, write ownership is acquired when the first attempted mutation triggers an atomic scope promotion and re-admission against the current active set of intents. The architecture employs brokered execution to bind capabilities to specific intent versions, leases, OS-level worktree locks, monotonic fencing tokens, and Git-tree provenance. Integration within Claim Plane involves verifying immutable patches and evidence.
Findings
A preliminary mechanism check was conducted using six pairs on CooperBench to evaluate the feasibility of Claim Plane. The results indicated:
- Static Claim Plane achieved six out of six pair passes with full serialization.
- Dynamic scope Claim Plane maintained parallel admission on half of the pairs.
- Dynamic scope Claim Plane performed seven successful scope promotions.
- Dynamic scope Claim Plane failed closed on two undeclared mutations.
The sample size used for this preliminary assessment was intentionally small and insufficient for drawing comparative claims. The findings are reported solely as evidence of feasibility for the Claim Plane mechanism.
Why This Matters
The Claim Plane architecture suggests a foundation for coordinating parallel coding agents by separating probabilistic planning from deterministic authority. This approach could potentially enable the development of future learned semantic-dependency models and facilitate frontier-model escalation specifically for unresolved cases.