Opportunities › Local AI agent orchestration platform › Proposed app
Benchwork
Run several AI code agents on your own machines, with your own API keys
- Score
- 55
- confidence 51
- Evidence
- 9
- items on the page
- Rivals
- 3
- named in sources
- First version
- 45 d
- estimate
- Type
- Desktop app
- PWA: runs in the browser, installs on Mac and Windows
- Market since
- 2025
- New market · AI estimate
ScreensConcept
Desktop app: Developers assign code tasks, review agent changes, compare results and manage spend—sustained oversight work done from a desk with keyboard input. Sample screens on a computer, in the proposed brand. Their numbers are sample data.
Active runs
1 waiting for approval4 running
About this appAI proposal
Benchwork is a control room for AI code agents that run on your laptop, in your containers or in sandboxes you manage. You assign tasks to several agents, such as Claude Code or Codex, watch them work and review what they changed, all from one browser tab.
It uses the API subscriptions and keys your team already pays for, so there is no second agent-platform bill. Agent code and repositories stay on machines you control; the hosted part only holds team settings, task history and audit records.
Built for engineering teams who want multi-agent workflows with limits, approvals and a clear record of who ran what, without tying themselves to a single vendor.
- Tagline
- Your agents, your machines, your keys.
- Voice
- PracticalDirectCalm
- Why this name
- A workbench is where tools are laid out and work stays in your own shop. The name suggests hands-on control of several tools at once.
In the first version
Members say which features they would use. It tells PWAS what people want; it changes no score.
Local runners
Install a small runner on a laptop, container or sandbox. Agents execute there; your code never has to move to a cloud service.
No member yet
Bring your own keys
Connect existing Claude, Gemini or Codex access. Benchwork orchestrates; you keep paying your model providers directly.
No member yet
Task board for agents
Split work into tasks, assign each to an agent and see status, branch and diff side by side in one view.
No member yet
Approval gates
Require a human to approve commands, file writes outside the repo or merges before an agent continues.
No member yet
Brand identityAI proposal
Check the name against trademark registers before using it.
Palette
#1F4D3FWorkshop Green · primary#F6F1E7Sawdust · surface#1E2421Charcoal · text#C9892BBrass · accent#4A6A86Slate Blue · data#B3402EBrick · alert
Typography
Aa
Benchwork
PricingAI proposal
The material gives no competitor prices, only that Dots, Bots and Muses charge separately. These tiers are a starting point to test, not a market-based figure.
Solo
Free
- 1 runner
- 2 concurrent agents
- 7-day history
Team
$15 a month
- Per seat
- Unlimited runners
- Approval gates
- Spend limits
- 90-day audit trail
Business
$39 a month
- Per seat
- SSO
- Audit export
- Retention policies
- Priority support
Would you pay Team · $15 a month?
Your answer tells PWAS which prices people accept. It changes no score.
No answers yet.
What it could be worthEstimate
What a software business making this much sells for today, from real deals. Pick a monthly revenue.
If it makes
Worth about
$194,000
Sale range $132,000 to $263,000
$120,000 a year in revenue
From 87 software businesses listed and sold on Empire Flippers in the last year, revenue vetted by the broker: they sell for about 19× their monthly revenue, most between 13× and 26×.
Measured on profit, the middle deal is 32× monthly net profit: about 2.7 years of profit.
An estimate, not an offer or financial advice. A real price depends on profit, growth, churn, how much of the work is the owner's, and the buyer.
How it would growAI proposal
How the product would reach new customers through its own use, and how the first ones would be found. A plan to test, not a result.
- Built-in loop
- Users invite others. A team lead invites teammates to share runners, task boards and approval queues. Reviewers are invited to approve agent steps and see results, so each team spreads use among its own members.
- Do they become customers?
- Invited teammates are developers who can become seat-paying users. Test whether invited reviewers on a free tier convert to paid seats once a team needs approvals and audit.
- First customers
- Engineering leads at software companies of 5 to 30 developers who already pay for Claude Code or Codex and want shared control over agent use.
- Where to reach them
- GitHub communities: Engage in discussions of local agent orchestration projects and offer a team layer that works alongside them.
- Developer forums: Post practical write-ups on running several agents safely with approvals and spend caps.
- Direct outreach: Message engineering leads who publicly discuss multi-agent workflows and offer a pilot.
- First milestone
- Five teams of at least five developers each pay for Team seats within 60 days of the first release.
Why people already pay for this
- VerifiedGitHub repository gained 8586 stars in 130 days
- Public claimExisting commercial platforms (Dots, Bots, Muses) charge separately for agent orchestration services, creating friction for cost-conscious teams.
- Public claimDevelopers want to avoid separate subscriptions and reuse existing API subscriptions for agent orchestration.
- Public claimThere is demand for local and self-hosted agent orchestration rather than cloud-only solutions.
- Verified
Tech stack
Every proposal builds on the same PWA base, installable from the browser on any phone or computer. Dashed boxes are what this app adds.
- AppNext.js and TypeScript
- PWAManifest and service worker
- UITailwind CSS
- Data and authSupabase Postgres
- PaymentsStripe
- EmailResend
- HostingVercel
- This app · Local runnerSmall cross-platform agent daemonRuns agents on the user's machine or container and connects outbound to the web app.
- This app · RealtimeSupabase Realtime channelsStreams run status and approval requests between runners and the PWA.
RisksAI inference
- CommercialA free open-source project in this space is popular, so teams may self-host it and not pay. The paid layer must offer clear team value such as approvals and audit.
- PlatformAgent vendors may change CLI behaviour, terms or key usage rules, which could break integrations or limit use of existing subscriptions.
- TechnicalRunning agents safely on laptops needs solid sandboxing and a secure runner connection across many operating systems.
- DataAudit logs may contain prompts and code snippets. Teams will want clear control over what is stored and for how long.
- AIParallel agents can produce conflicting changes or run up token costs, so limits and review flows must work from day one.
Everything marked AI proposal or AI inference was written by a model from the evidence on the opportunity page. The proposed name must be checked against trademark registers before it is used.