Agent fleet
Coordinate planning, implementation, testing, and independent review across the models and harnesses you choose.
Waymark Studio · In active development
Mission control for agentic software delivery.
A native macOS workspace for turning issues into isolated agent runs, checked changes, intentional approvals, and recoverable releases.
Overview
Product Lab
One decision needs youA release is ready for review
The exact production plan and validation evidence are waiting.
Live workOpen Work
Harden release recoveryProduct API
RunningRefresh product readinessWaymark Studio
QueuedReview field sync evidenceiOS Client
QueuedNeeds you1
Review production plan
Artifact, target, checks, and rollback proof are ready.
Capacity and readinessOpen Settings
Waymark Engine Ready
Repository proof Ready
Agent harnesses Available
Work queue 3 waiting
01Studio
The native Mac control room for work, Products, evidence, and decisions.
02Waymark Engine
The background authority that owns durable state and keeps governed work moving.
03
waymark CLIA headless peer of the GUI for scripts, automation, and exact machine-readable state.
One operational picture
From backlog to delivery without losing the thread.
Studio keeps product scope, agent work, repository gates, approvals, and release evidence attached to the same task.
Managed workspaces
Each run starts in a Broker-owned worktree with explicit scope and recovery protection.
RepoSeatbelt
A separate agent-proof recovery path protects local source before risky changes and integration writes.
Validation profiles
Map repository expectations into repeatable local and disposable-runner checks with visible evidence.
Approval policy
Stop after merge, require an exact production approval, or automate reversible delivery under saved policy.
Fail-closed control
Emergency stop, stale proof, scope drift, unsafe commands, and incomplete cleanup all halt execution.
From issue to production
Give Studio the work.
Choose how far it goes.
GitHub Issues is the first task source. More trackers are planned. From there, Studio coordinates the agents, proof, privileged actions, and human decisions around one task.
Bring in real work
An issue becomes a governed task.
Requirements, non-goals, repository scope, risk, and finish mode are captured before any agent touches code.
GitHub Issues · available firstHarden release recovery
Scoped · reviewed · ready
Other task sources planned
Fan out the work
Plan. Build. Test. Review.
Different agents can take different roles. Independent reviewers challenge the implementation before it moves forward.
Keep work isolated
Every agent gets its own lane.
Broker-managed worktrees keep changes separate until they are reviewed and ready to integrate. Bounded multi-run scheduling is in active development.
PlanBuildTestReviewTask
Prove the change
Review and tests must agree.
Local checks, security scans, disposable runners, UI evidence, and reviewer findings resolve into one visible result.
Independent reviewPassedRepository checksPassedSecurity scanPassedRelease verificationPassed
Broker-owned delivery
Each boundary has its own policyPR. Merge. Production.
Studio can create the commit, push the branch, open the PR, merge approved work, and run the exact authorized release path.
Capability stays outside the model
Agents never receive the keys.
Repository tokens, cloud credentials, signing identities, servers, and production controls stay behind the native Broker wall.
Agents do the creative work. The harness controls what they can reach—and what they can never reach.
Automation is your choice
Step by step. Or policy by policy.
Pause at every decision while you build confidence. Then automate the routine paths you trust without removing the hard safety gates.
Speed changes. The security boundary does not.Review every step
Approve the plan, agent run, PR, merge, and release separately.
Automate agent work
Let approved agents plan, build, test, and review while delivery waits.
Automate through merge
Allow verified, policy-matching changes to open and merge PRs.
Automate delivery
Run an exact reversible release path under saved product policy.
Local-first
Your code, credentials, and agent subscriptions stay yours.
Studio uses installed agent harnesses and execution on machines you own. It does not need to become a hosted IDE or token marketplace to coordinate serious work.
- Native macOS control plane
- Your choice of models
- Local Tier-B source and transcripts
- Disposable runners you own
Agent request
Local BrokerApproved action
Credentials never cross into the agent process.Your agents, one harness
Use the right model for each role.
Studio is built around installed agent harnesses and your own model choices—not token resale or a locked-in model cloud.
CodexClaude CodeLocal modelsCustom adapters
Integration depth varies by harness while Waymark is in active development.
Built for operators
Five places.
One operational picture.
The current Studio workspace keeps navigation quiet and puts each decision where it belongs.
Health and attention
See the exact Lab state, current work, decisions, capacity, and recent activity.
Queue and active runs
Inspect the live run, queued work, blockers, recovery, and approval gates.
Repos, rules, and setup
Keep repository readiness, task sources, models, and delivery policy together.
Activity and evidence
Follow durable outcomes without turning raw source or transcripts into telemetry.
Lab, agents, and app
Manage the local Engine, installed harnesses, credentials, updates, and capacity.
In active development
Talk with me about Waymark.
Waymark is in active development. If your organization is working out how to let AI agents do real work safely, I'd like to hear about it, whether to compare notes or to discuss Waymark itself.