DevSpec

The home base for AI-native software teams

Your team, your codebase, and every AI tool you build with — working from one source of truth.

DevSpec is the shared workspace where your team plans the work, your AI agents — in Claude Code, Cursor, or anything that speaks MCP — pick it up and ship it, and every branch, decision, and result stays on the record. One home for the whole loop, instead of a dozen disconnected tools.

Start Your Project See how the loop works

Free trial · 15,000 credits to start

payments-api — team session

Ali, Priya, Tom + DevSpec

3 online

Priya

Stripe webhooks keep timing out on retries. Can we just stage a fix for the backoff and get it out today?

The retry loop is in lib/payments/webhooks.ts:42— fixed 2s delay, no jitter. You decided in May that webhooks verify out-of-band, so the fix is safe to stage. I've written the acceptance criteria; rush it to the top?

#214

Exponential backoff for Stripe webhook retries3 acceptance criteria · staged for Autopilot

staged

22 minutes later

Done. Pushed as [devspec:214], deployed to staging, and the AI test walked all three criteria — passing, on video. Want the link?

— a session, start to shipped. The duck doesn't just listen.

01The Loop

From decision to verified code.

Most software decisions happen in conversation — then someone has to turn them into tickets and hope nothing gets lost. DevSpec carries the decision all the way to a deployed, tested change.

DiscussA team session with AI that knows your code, schema, and history.

CapturedDecisions, conventions, and risks recorded as durable memory.

Work definedAction items with acceptance criteria — nobody writes a ticket.

Autopilot runsRunners on your machines implement in isolated worktrees, in parallel.

Deployed & verifiedDeploys trace back to the work; AI browser tests prove it works.

Memory compoundsEverything learned feeds the next loop. Sharper every week.

Knowledge feeds back into the next conversation

02Versus the Market

Stop engineering the harness.

Most tools hand you a powerful agent and leave you to wire up its context, memory, and access by hand. DevSpec ships the harness already built — purpose-built for the AI-Driven Development Lifecycle.

The Alternatives

DIY Harness Engineering

Bolted together by hand. Maintained forever.

DevSpec

The Managed Harness

Purpose-built for the AI-DLC.

Context

Hand-written .cursorrules and CLAUDE.md files that drift out of date the moment the codebase moves — context rot you maintain forever.

Context

A Shared Brain streams the right context, decisions, and conventions to your agent every session. No rule files to write or babysit.

Integration

Custom MCP servers duct-taped together to bridge your editor to Jira or Linear — one more brittle service to host and debug.

Integration

Planning, memory, and the work queue are native and arrive over one MCP connection. Nothing to build, nothing to maintain.

The lifecycle

Expiring auth tokens and amnesia-prone terminal sessions that forget what you decided five minutes ago.

The lifecycle

Purpose-built for the AI-DLC: Inception is handled natively, and "AI executes, human validates" is the default — not an afterthought.

See the full comparison

03The Difference

What makes it different

Not a chatbot bolted onto a board. The conversation, the record, the execution, and the proof are one system.

It has actually read your repo

Code, schema & memory

Indexed code across seven languages, live database schema, and twelve types of recorded memory. Ask why something is the way it is, and you get the decision, the date, and the file — not a plausible guess. New teammates onboard from the record, not from interrupting seniors.

Session 23 · payments-apiMemory recall

Why don't we verify webhook signatures inline? Feels wrong.

You and Tom decided this on 12 May — inline verification blocked the worker pool under load, so verification moved out-of-band. The trade-off is recorded; the constraint was p95 latency, not security. If the load profile has changed, it's worth revisiting — but that's the history.

decision

Webhook signatures verify out-of-bandRecorded 12 May · session 14 · confidence 0.9

It holds the thread — even the ones you dropped.

Conversations become shipped code

Autopilot on your machines

Decisions become action items with acceptance criteria, and runners on your team's machines pick them up — claiming atomically, implementing in isolated worktrees, committing with traceable tags. What used to take a sprint takes a conversation.

Autopilot — queue2 runners active

#214

Exponential backoff for Stripe webhook retries

runner mini-pc-01 · worktree wt-3f2a

implementing

#209

Empty state for the deployments page

pushed · [devspec:209] → staging

deployed

#216

Rate-limit headers on the public API

staged behind #214 — shares lib/api/middleware.ts

staged

The queue, mid-afternoon. Nobody is typing.

Deployed isn't done — verified is

Evidence, not optimism

Every deployment links back to the conversation that caused it. Then an AI browser test walks the acceptance criteria on the live app and brings back video. Failures become bug items with the evidence attached, ready to queue.

Test run — #214staging

Exponential backoff for Stripe webhook retries

AI browser test · video 0:48 · 11 screenshots

Trigger a webhook retry from the Stripe test dashboard
Retries arrive with exponential backoff and jitter
Event marked processed exactly once in the admin view

confidence 0.92pass — confirmed

Ten minutes after deploy, the proof arrives on its own.

04An Index

Six pillars, one loop.

01Team SessionsYour whole team, one AI conversationMulti-user AI sessions with voice, @mentions, polls, and plans — backed by full project context, so nobody explains the codebase twice.Read more →

02Code IntelligenceAI that has actually read your repoIndexed code across 7 languages, live database schema awareness, and a memory that keeps every decision your team has ever made.Read more →

03PlanningTracked work without writing ticketsConversations become action items with acceptance criteria automatically. Goals track real progress from the work that ships — no separate tracker, no backlog to groom.Read more →

04AutopilotQueue work. Agents ship it.Coding agents on your machines pick up staged action items, implement them in isolated worktrees, and push — while you do something else.Read more →

05VerificationEvery deploy traced and testedDeployments link back to the conversations and commits that shipped them — and AI browser testing checks the result actually works, on video.Read more →

06MCP & IntegrationsYour IDE inherits everything33 MCP tools give Claude Code, Cursor, and any MCP client your full project context — plus GitHub and Slack built in.Read more →

05Three Seats at the Table

The entire team, in one place.

Technical and non-technical collaborators work with the same AI, in the same thread, against the same record. No one is guessing.

I shape the product

Describe it. Watch it ship.

Speak naturally — voice or text, no technical language required. DevSpec remembers every decision and conversation, and your plans flow directly to the people (and agents) who build.

Team Sessions→

I build the software

Context arrives before you do.

Your IDE already knows what was decided, why, and what to pick up next — over MCP. Every commit, convention, and rationale is on record and searchable.

MCP & Integrations→

I verify the work

Evidence, not optimism.

Every change arrives with what was done, why, and how to test it. AI browser tests walk the acceptance criteria on the deployed app and bring back video.

Verification→

06The Ecosystem

Plugged into your whole stack.

Repositories, hosting, databases, observability, and the coding agents your team already uses — connected through one open protocol.

Repositories

GitHubGitLabBitbucketGitea

Hosting & Deploys

VercelNetlifyRailwayRenderCoolify

Databases

SupabaseMongoDB Atlas

Logs & Observability

AxiomLoki

Coding Agents

Claude CodeCodexCursorGrok BuildAntigravity

See all integrations

Plainly

0MCP tools for your IDE

0Languages indexed

0Types of project memory

0 minTo a first session

Your code stays yours

DevSpec reads your repositories via the GitHub API and never trains on your data. The agents that write code run on your machines, under your own accounts.

Every change traceable

All knowledge is structured, searchable, and attributed — who said it, when, and why. The full trail runs from discussion to deployed, verified code.

Honest pricing

$30 per seat / month, 30,000 credits included. No hidden costs, no 'contact sales' gates, and unused credits roll over.

Works with Claude Code and Cursor — and any MCP client.

The workflow that didn't exist. Until now.

Set up your first project in under five minutes. Connect your repo, start a session, and see what your duck can do.

Start Your Project

14-day free trial · 1,500 credits included