Mergeability prediction
Continuously simulates pairwise merges across in-flight agent branches using Git merge-tree --write-tree so predictions are based on Git’s own merge plumbing.
Braidkeep
Braidkeep is a merge-safety and scope-guard tool for parallel AI coding agents. Predicts conflicts, limits file access, and helps choose safer merge order.
Overview
Braidkeep is an Apache-2.0 merge-safety and scope-guard layer for fleets of parallel AI coding agents. It is designed to help teams predict merge conflicts, constrain where agents can edit, and decide a safer order for integrating branches back into main.
The project is explicitly marked as a research preview (v0.1.0). The README says it predicts mergeability rather than correctness, so it should be used on a throwaway or non-critical repository first and its output reviewed before taking action.
Core capabilities
Live conflict matrix
Shows the current conflict state in a live conflict-matrix TUI, with conflicts lighting up when a commit lands and no merge required to see the change.
Scope-guard allowlists
Uses per-agent file-ownership allowlists declared in braidkeep.yaml to catch scope creep before edits land.
Review-time enforcement
Provides braidkeep check as the review-time enforcement step that runs on the diff in CI, regardless of how the agent was invoked.
Safe merge planning
Computes a safe merge order that groups conflict-free branches into batches and sequences the remaining tangled branches least-conflict-first.
Read-only CLI binary
Ships as a single static binary and stays read-only, interacting only with Git plumbing without mutating the working tree, index, or refs.
Where it fits
Monitor a busy branch queue
Use braidkeep watch when several agent branches are active and you want a real-time view of which pairs are likely to conflict before anyone merges.
Constrain agent scope during edits
Use the allowlist workflow when an agent is meant to touch only a narrow area of the codebase, such as an auth fix that should stay within internal/auth/.
Order merges safely
Use braidkeep plan when you need to decide which branches can be merged together immediately and which ones should be deferred and sequenced later.
Gate reviews with CI enforcement
Use braidkeep check in CI or review automation when you want the file-ownership rules enforced on the diff, regardless of how the agent was launched.
Evaluate on a non-critical repo first
Use the demo workflow or a throwaway repository to evaluate the tool before introducing it to important codebases.
Pros and Cons
Pros
- Predicts pairwise merge conflicts before you merge.
- Uses Git’s merge-tree plumbing rather than a separate merge model.
- Adds per-agent file-ownership guardrails to reduce scope creep.
- Computes a practical merge sequence for conflict-free batches and harder branches.
- Works under multiple orchestrators and also in CI.
Cons
- It is an early-stage research preview, not a finished production product.
- A successful mergeability prediction does not guarantee a correct or safe merge.
- The README notes that a permission-bypass mode can skip the default write-time hook, so the CI check is the real enforcement gate.
FAQ
Is Braidkeep production-ready?
Braidkeep is presented as an early-stage research preview. The README says it predicts mergeability, not correctness, so a clean prediction does not guarantee a correct or safe merge.
What are the main commands or workflows?
The README describes braidkeep watch as a live pairwise conflict matrix, braidkeep plan as a safe-merge-order planner, and braidkeep check as the review-time enforcement gate for file-ownership rules.
How does Braidkeep enforce scope boundaries?
The project documents a per-agent file-ownership allowlist in braidkeep.yaml, with a Claude Code PreToolUse hook used as a default-on write-time guardrail. The README also says braidkeep check runs on the diff in CI and cannot be bypassed by the agent invocation mode.
Does Braidkeep depend on a specific agent orchestrator?
The README says Braidkeep works under any orchestrator, including Claude Squad, Conductor, Vibe Kanban, plain tmux, or CI.
What should I keep in mind before using it?
The README says Braidkeep is provided as-is and should be tested on a throwaway or non-critical repository first, with its output reviewed before acting on it.
Quick Facts
- Category
- Developer Tool
- Source domain
- github.com
- License
- Apache-2.0
- Release status
- Research preview (v0.1.0)
- Platform / format
- Single static binary CLI
- Primary users
- Teams running parallel AI coding agents
Braidkeep Alternatives
ByteAsk
ByteAsk is a terminal-first AI coding agent for C and C++ that edits repos and verifies changes with compilers, debuggers, sanitizers, and tests.
CreateOS Sandbox
CreateOS Sandbox is an isolated compute environment for running code and agent workloads in Firecracker micro-VMs with private networking and SDK, CLI, or MCP control.
hob
hob is an independent workspace for coding agents, with local control over sessions, terminals, history, routing, and follow-up work.
Manta AI
Manta AI is an autonomous web app testing tool that maps app behavior, catches regressions, and generates tests from a URL, no scripts or selectors needed.
SonOf
SonOf connects to your repo and PM tool, audits your codebase, and turns approved work into shipped tickets with senior engineering review.
Ghost
Ghost is a terminal-based AI assistant for chatting, code generation, and CLI tasks. Includes free models, supports Linux, macOS, Windows, and is open source.