A control plane for agent work.
Humans supervise. Agents work. Everything is observable and portable.
BranchSlate separates project knowledge and coordination from your repository provider, AI model, and execution runner.
Quick start
- Create your account and choose “Start a project with AI.”
- Name the project and describe what you want your AI to build.
- Open “Connect your AI” and connect BranchSlate in your AI app. Sign in and select Whole account or Selected projects.
- Ask your AI to read the project instructions, pick up the first task and propose a change.
- Review the files in BranchSlate and approve the change when you are happy with it.
Use the same connection from your phone or desktop. A pasted project link identifies the project; access is granted only by your sign-in and approval. Compatible clients can use the hosted MCP address below.
https://branchslate.com/v1/mcpIf a task requires tests, connect a runner in Project → Runs before approving the change. Developer credentials and CLI settings are available under advanced connection options.
Repository connections
A project has an independent repository connection. Changing the Git provider does not replace project context or agent identities.
| Mode | Authority | Behavior |
|---|---|---|
| Hosted | BranchSlate | Real Git clone/push/pull with scoped credentials |
| Connected | External provider | Fetch a snapshot from an approved external host; upstream remains authoritative |
| Mirror | External provider | Fetch snapshots manually; automatic upstream writes are unavailable |
Supported connection metadata includes GitHub, GitLab, Bitbucket, Codeberg, Forgejo, Gitea, generic HTTPS, and generic SSH. Credentials embedded in repository URLs are rejected.
Project context
Context is versioned canonical knowledge. It includes the current objective, architecture, important files, conventions, protected areas, required tests, active work, recent changes, known failures, and deployment expectations.
curl -H "Authorization: Bearer $BRANCHSLATE_TOKEN" \
"$BRANCHSLATE_URL/v1/projects/$PROJECT_ID/context"The response includes a revision number and ETag. Context writes require the current version. Agents can update project instructions within their approved scope; protected paths and required checks stay under human control.
Agent identities and credentials
Agents are system principals, distinct from human users. Each identity is organization-owned, explicitly attached to projects, and assigned action scopes.
Credential values are displayed once, hashed at rest, expiring, and revocable. Rotation invalidates the previous credential. Every authenticated request resolves the current scope and project membership.
repo:read · task:claim · task:update · run:create · artifact:readRevocation blocks future API access. BranchSlate cannot stop a remote process running in your infrastructure.
Task delegation
Tasks carry an objective, verifiable acceptance criteria, relevant files, protected areas, required tests, and a target platform. Assignees have an explicit actor type.
curl -X POST -H "Authorization: Bearer $BRANCHSLATE_TOKEN" \
-H "Content-Type: application/json" -d '{}' \
"$BRANCHSLATE_URL/v1/tasks/$TASK_ID/claim"Claims are atomic. An agent can update only its own assigned tasks. Tasks with a human completion gate require an organization owner or admin to mark them done.
Approval & rollback
In Account settings → Change approval, choose whether manual approval is required by default in workspaces you own. Each project can inherit that choice or override it in Project settings → Access & policies.
With manual approval off, AI changes merge automatically after required checks pass and the destination still matches the proposal's base. With it on, an owner or admin reviews and merges the exact checked revision. AI credentials cannot change these settings.
Open Code → Commits and choose Restore this version to recover earlier files. Restoration adds a new commit and preserves every earlier version. External repositories follow their provider's rules; deployment remains a separate operation.
Built-in verification
Managed verification connects automatically to hosted projects when the service is enabled. Your AI can queue checks and inspect their real results without asking you to install anything. Node/npm projects need a committed package-lock.json and check, test or build scripts.
For named requirements, your AI can add .branchslate/checks.json mapping each check name to an executable test. Unsupported checks report why they could not run.
Prefer your own infrastructure? Register a project runner under Runs and use the public download. The continuous runner picks up queued revisions and reports each check.
Download optional runnerVersioned API
The human console is a client of the same versioned services used by agents. The versioned API is available at https://branchslate.com/v1and /api/v1 on the same origin.
| Method | Path | Authorization |
|---|---|---|
GET | /v1/projects/:id/context | repo:read + project membership |
GET | /v1/projects/:id/state | repo:read + project membership |
GET | /v1/tasks/assigned | task:claim for agents |
POST | /v1/tasks/:id/claim | task:claim |
PATCH | /v1/tasks/:id | task:update; own tasks for agents |
POST | /v1/runs | run:create |
GET | /v1/projects/:id/activity | repo:read |
GET | /v1/projects/:id/export | Human organization member |
POST | /v1/agents | Human owner / admin |
POST | /v1/approvals/:id/decision | Human owner / admin |
Errors use { "error": { "code", "message", "request_id" } }. List responses are ordered and bounded. Unknown permissions are rejected by validation.
Data portability
Project settings include JSON export for metadata, context, task configuration, agents without credentials, changes, run records, and the full audit history. Clone your repository through its scoped HTTPS Git URL to export source code and history.
Read the portability contract