AI & agents

A team of agents that know your company.

Each agent reads the project it's working in: purpose, milestone, open tasks and docs. They run on open models, and you pay per token with no seat licences.

● checking models…Agents 6Output cap 4,096 tokensLocal

Workflow

One milestone, three agents, real output.

This is the flow the app runs: the project's context goes to each agent, and their work lands back in the project as tasks, docs and a live site.

Project · Northwind API

Usage-based billing for indie SaaS teams.

purpose6 tasks2 docs

◎Milestone

Private beta with event ingestion, invoices and a dashboard.

◇Product Lead

Writes the one-page spec and plans the milestone.

gemma4:12b

</>Staff Engineer

Writes the ingestion endpoint and the landing page.

qwen2.5-coder

?Support Lead

Drafts the beta FAQ and onboarding email.

gemma4:12b

✓+6 tasks on the board

With acceptance criteria.

☁Site deployed

/s/northwind · $0.0004

¶Docs saved

Spec and FAQ in the project.

The agents

Six roles, ready on day one.

</>code model

Staff Engineer

Production code, reviews, architecture and trade-offs. Prefers code over prose.

◇general model

Product Lead

Specs with scope, non-goals, milestones and acceptance criteria.

?general model

Support Lead

Customer replies and help articles, and never invents a policy.

↗general model

Growth Marketer

Developer-facing landing copy, launch posts and emails, with variants.

$general model

Ops & Finance

Budgets, pricing models and runway, with the arithmetic shown.

◎general model

Chief of Staff

Weekly plans, decision records and investor updates.

Ask anywhere

Press ⌘K. Ask about anything.

The command bar knows which page and project you're on. Answers stream in, show their cost, and continue as a full conversation when you need more.

⌘K · Northwind API

✦ ask what is left for the beta milestone?

Four of six tasks are open. The riskiest is the meter
logic for usage calculation: it blocks invoices and
the dashboard. Ship it first, then invoices.

gemma4:12b · 191 tokens · $0.000061 · Continue in chat →

Plan with AI

Structured output, straight onto the board.

Planning asks the model for JSON that matches a schema, so every task arrives with a title and acceptance notes, ready to move across the board.

// model output (schema-constrained)
{ "tasks": [
  { "title": "Implement event ingestion endpoint",
    "notes": "POST receives raw usage events. Acceptance: validates payload and stores events." },
  { "title": "Build usage dashboard UI",
    "notes": "Real-time usage per customer. Acceptance: shows data from the billing API." }
] }

Models

Open models, per million tokens.

Models marked cloud are free third-party models through OpenRouter, and their providers may log prompts.

Model	Best for	Input / 1M	Output / 1M
Loading models…

Usage

See where the tokens go.

Billing breaks usage down by model and by agent. The example below shows the shape of it.

Staff Engineerexample

code model · most active Tue–Thu

Support Leadexample

general model · steady daily

Growth Marketerexample

general model · launch weeks

Put the agents to work.

Start free API reference