platformREADY
storefrontREADY
developer-apiREADY
Start with the right setup.
Keep environments, project instructions, and tools together so each run starts with the context it needs.
MANAGED AGENT OPERATIONS
Your agents run.
Your team ships.
One operating layer for your agents, workflows and model spend. Pre-buy usage at negotiated rates or pay after consumption, while your team stays focused on shipping.
$600 / month5 developers · 10 repositories · model usage billed separately
Dependency maintenanceExample workflow
storefront / update-dependencies
01
Start with the right environment
Repository, tools and project instructions.
02
Run the agreed workflow
Approved patch updates. A bounded task.
03
Verify before handing off
Lockfile integrity and regression checks.
Ready for your review
Changes, checks and context in one handoff.
YOUR AGENTS.
WORKING TOGETHER.
Claude Code
Codex
OpenHands
GitHub
YOUR DEVELOPMENT STACK01 / THE CONTROL PLANE
A clear view.
From task to pull request.
See what is running, where it is running, and what needs your attention. Keep the context of every handoff.
Work is moving.
You have the controls.
CONNECTED REPOS08/ 10
WORKFLOWS03configured
REVIEW POLICY Human approval
ACTIVE WORKFLOWSInspect a workflow
fix/auth-typecheckInspect workflow below
THE WORK AROUND THE WORK
Your agents write code.
Give them a system.
Every extra agent brings another environment, handoff, and decision. Build a repeatable operation around them.
01
02
Make the handoffs repeatable.
Turn recurring engineering work into workflows with defined steps, checks, and a clear finish line.
Merge to mainApproval required
Your team has the final say.
Keep judgment in your hands.
Define permissions and review boundaries. Get a traceable result your team can inspect before it ships.
02 / A REPEATABLE ROUTE
A task goes in.
A reviewable
result comes out.
Give every run a destination and a set of checks. Follow an example from the first instruction to the final handoff.
acme / platformWORKFLOW RUN
Define the work01
Resolve the auth type error
Reproduce the CI failure on the latest auth change, make a targeted fix, and run the full check suite.
Run in an isolated environmentClaude Code
Project context and tools attached.
Check the resultPOLICY
Typecheck, unit tests, and lint
Hand off for human review04
src/auth/session.ts
+ const session = await getSession(); 3 checks passed · Ready for review Nothing merges without your approval.
03 / BUILT AROUND YOUR BOUNDARIES
Autonomy needs
a set of controls.
Set the rules once. Apply them across the work your team delegates.
harness.configWORKSPACE POLICY
01
02
03
04
05
06
07
08
09
10
02
03
04
05
06
07
08
09
10
# A clear finish line for every run workflow: fix-failing-build checks: - typecheck - unit_tests - lint merge: human_approval
A reviewable PR is the destination.
MODEL CREDITS / PART OF THE OPERATION
Fund the work.
Let the harness meter it.
Pre-purchase usage at an agreed bundle rate, or pay after consumption at a higher rate. One spending view across your approved workflows.
Prepaid. Better rates.
Buy service credits through Paddle. Eligible models use the negotiated rate card, and every run draws down your shared balance.
Postpaid. More flexibility.
Keep cash available until usage is billed. Pay the standard token rate plus 10%, within an agreed monthly ceiling.
Your keys remain an option.
Use your own provider accounts where required. Direct-provider usage stays outside the HarpessOps credit balance.
Workspace creditsInteractive example
Available credits$500.00
Prepaidstorefront / maintenanceProject budget
$0.00 / $34.00Credit purchase+$500.00
Replay a run to see usage deducted from the shared credit balance.
0 completed runs. Other projects keep their own spending limits.
A $10 standard-rate run costs $8.50 at the example 15% prepaid discount. This demo makes no model calls or charges.
04 / ONE TEAM. ONE PLAN.
For the small team
doing a lot.
A shared operating layer for the people building, reviewing, and shipping with AI.
Prepay for model usage.
Or choose postpaid billing.
Or choose postpaid billing.
MANAGED TEAMMONTHLY PLAN
$600
/ monthUSD · per team
An operating layer, with recurring maintenance in scope.
Up to 5 developers
Up to 10 repositories
3 agreed workflows
2 hours guided setup
2 maintenance hours / month
Weekly workflow health summary
Add prepaid model credits, approved postpaid usage, or your own provider keys. The $600 subscription does not include model consumption or execution infrastructure. Billing through Paddle.
EXPLORE THE POSSIBILITIES
Explore the cases Built around the way your team works.
Explore workflows for senior developers, product teams, and software studios.
A FEW THINGS TO KNOW
Good questions.
What exactly is a harness?
A harness is the operating layer around an AI agent: its environment, tools, instructions, permissions, checkpoints, and completion checks. HarpessOps brings those pieces together across your team's projects.
Do we keep our existing agents?
Yes. HarpessOps is built around your existing agent stack. We work with you to define the agents, repositories, tools, and workflows in your workspace.
What does the $600 cover?
The managed team plan covers up to 5 developers, 10 repositories, and 3 recurring workflows, with guided setup and monthly maintenance. Model usage is separate: pre-purchase HarpessOps credits at an approved bundle rate, use postpaid at standard rates plus 10%, or keep supported provider keys. Execution infrastructure is separate. Visit Pricing for the full scope.
Can I pre-purchase model usage?
Yes. Choose a credit pack after approving the model rate card. Your harness deducts metered consumption from the balance. Credits fund eligible HarpessOps inference; Paddle handles the purchase. Postpaid is available by agreement at a higher token rate.
Who approves code changes?
Your team defines acceptance checks and review boundaries. Workflows hand off a pull request with the context and evidence your engineers need to review the change.
How do we get started?
Request a demo and share your current stack. We agree on the first workflows, confirm access boundaries, and establish a baseline before onboarding your team.
LET'S BUILD A BETTER WAY TO WORK
Your agents have potential.
Give them an operation.
Request a demo. Tell us what your team wants to hand off.