v0.1 · the first engineering workspace with a full AI team built in

Your AI engineering team,
shipping with you.

Turn a one-line idea into a reviewed PRD, a sequenced roadmap, and a board full of tasks — in an afternoon, not a quarter. The AI team drafts. You approve every gate.

// live runrunning
saarthi — team run
start-team --project "user auth redesign"
Product Managerintent captured11ms
AnalystPRD written · 14 pages38s
Project Manager5 milestones sequenced9s
Tech Lead23 tasks on board14s
Architectindexing codebase...
why now

Every AI tool covers
one slice.

You can ship a PRD here, a roadmap there, tasks in another app, and code answers somewhere else. The artifacts never talk. The context never compounds.

Devin · Factory
Execution only

No PM lifecycle. No artifacts. Code-only.

Cursor · Copilot Workspace
Code editing

Doesn't know your roadmap or your spec.

Linear AI · Jira AI
Task triage

Sorts work. Doesn't define it.

ChatPRD · Productboard AI
PRDs only

PRD ends, the workflow stops.

what we ship

One workspace.
The whole arc.

01PRD
02Roadmap
03Milestones
04Tasks
05Codebase context
06Approval gates

A full AI engineering team — product manager, analyst, project manager, tech lead, and codebase architect — working in your PRDs, your plans, your board. You approve every gate.

the flow

From a sentence
to a shipped roadmap.

One thought goes in. Five agents take it from there.

01Discovery
Product Manager
supervisor
Intent captured
02Spec
Analyst
agent
PRD saved
03Plan
Project Manager
supervisor
Milestones live
04Build
Tech Lead
agent
Board populated
05Ground
Architect
agent
Cited in code
how you'll work together

Built for how engineers
actually work.

Not a chat window. Same PRDs, same plans, same board your team already uses — with AI specialists drafting, and you in command at every gate.

01human
You
Drop a goal in chat
"We need notifications when a customer's plan is about to expire."
02agent
Product Manager
Captures intent, asks one clarifying question, hands off
Loads product context and routes to the Analyst.
03agent
Analyst
Drafts a complete PRD
Problem, goals, user stories, requirements, success metrics — saved.
04human
You
Edit inline, push back, approve
Tweak a goal, sharpen a metric. The agent edits surgically — no rewrites.
05agent
Project Manager
Generates milestones from the approved PRD
3–7 milestones, sequenced, dated, prioritized.
06human
You
Reshape, reorder, approve
Move a milestone, change a date. Approval is the gate to the next phase.
07agent
Tech Lead
Breaks each milestone into board-ready tasks
Acceptance criteria, effort estimates, priorities.
08together
You + Architect
Build, with the codebase on tap
Ask the Architect anything. It retrieves from your code and cites the file.

Agents do the drafting. You hold the gates.

Approval is what moves work to the next phase — never the agent's call.

the roster

Five specialists.
One coordinated team.

supervisor
Product Manager
Discovery + vision

Captures intent, asks the one clarifying question that matters, loads the product context, and hands the brief to the Analyst. The orchestrator that keeps every phase aligned.

Clarifies intentHolds contextPhase-gates
agent
Analyst
PRDs + requirements

Drafts a complete PRD — problem, goals, user stories, functional + non-functional requirements, success metrics. Edits surgically when you push back. Saves every time.

Writes PRDsEdits in placeAuto-saves
supervisor
Project Manager
Roadmap + milestones

Turns an approved PRD into 3–7 milestones — priorities, target dates, dependencies, ordered chronologically. Refuses to plan from an unapproved spec.

Milestone-sequencingDate-awareGate-keeps
agent
Tech Lead
Task breakdown

Decomposes each milestone into 3–7 concrete tasks with acceptance criteria, effort estimates, and priorities. Tasks land on your board, ready to assign.

Acceptance criteriaEffort estimatesBoard-ready
agent
Architect
Codebase expert

Knows your migrations, routes, tests, and architectural decisions. Never answers without retrieving. Always cites the file. Says "I don't know" when it doesn't.

Codebase-groundedCites sourcesNo bluffing
joining the team
Code Reviewer· PR feedbacksoon
Data Analyst· metrics + insightssoon
ML Specialist· model designsoon
QA Engineer· test coveragesoon
Security Auditor· threat reviewsoon
under the hood

Engineering culture,
built into the team.

Specs that get reviewed. Plans that have gates. Code answers with citations. Every output scored. The habits of a good engineering org, encoded into how each agent works.

01

Real tools, not just chat

Each agent has the tools it needs to do actual work — fetch PRDs, save plans, write tasks, query knowledge, search the web. Output is artifacts, not transcripts.

02

Workflows, not one-shot prompts

Every specialist runs a deterministic pipeline — gather, write, format, verify. Predictable, debuggable, improves over time.

03

Persistent memory

Agents remember what they've worked on. The Architect has its own memory of your codebase. Conversations pick up where they left off.

04

Every output is scored

Built-in scorers grade PRDs for completeness, roadmaps for coverage, tasks for clarity, delegation for accuracy. The team gets better, measurably.

05

Grounded in your knowledge

Upload docs and index your codebase. Agents retrieve, cite sources, and refuse to bluff. If it's not in the knowledge base, they say so.

06

Smart model routing

Lightweight chat uses a fast model. Heavy planning uses a deeper model with extended thinking. Speed where it matters, depth where it counts.

begin

Stop briefing.
Start shipping.

Open a workspace. Drop in a goal. Walk away. Come back to a PRD, a roadmap, and a board full of tasks — grounded in your context, scored for quality, ready for review.