MCP & Integrations

DevSpec isn't a replacement for your editor — it works alongside it. 33 MCP tools give Claude Code, Cursor, and any MCP-compatible client your project's full intelligence: decisions, conventions, action items, code index, and schema.

Start Your Project Read the docs

claude code — ~/payments-apimcp: devspec

> implement DevSpec item #214

⏺ devspec · reserve_work_items → #214 exponential backoff for webhook retries

⏺ devspec · get_decisions → 3 decisions on payments, incl. out-of-band verification

⏺ devspec · claim_work_item → claimed by this agent

Reading lib/payments/webhooks.ts — your May decision means I can change the retry loop without touching verification…

— Claude Code, opening with the context already in hand.

The Problem

Stop explaining your project twice

Your team decides something in a meeting, then each developer re-explains it to their coding assistant from memory. Context evaporates at every tool boundary — and the assistant that writes the code never saw the conversation that decided it.

After a DevSpec session, a developer opens Claude Code. It already knows the decision, the conventions, the open action item, and the schema — over MCP. Nothing copy-pasted, nothing lost in translation.

The Mechanics

How it works

One endpoint, and every tool your team uses starts working from the same brain.

Connect once

Generate a project token and add the DevSpec MCP endpoint to Claude Code, Cursor, or any MCP-compatible client. Setup is a config block, not an integration project.

Fig. i — MCP setupapp.devspec.ai/settings

Adding the DevSpec MCP endpoint to a coding tool

A config block, not an integration project.

Read everything

Your tools can search the code index, read action items and acceptance criteria, recall team memories and decisions, browse generated docs, inspect the database schema, and pull testing briefs — 24 read tools in all.

Fig. ii — Read toolsclaude code · cursor · any mcp client

An external agent searching the DevSpec code index and recalling decisions

The index your sessions use is the index your IDE queries.

Write back

For explicitly requested items, agents call reserve_work_items then claim_work_item, link commits, and record implementation evidence. The project record updates while code is written — without a work router.

Fig. iii — Write-backapp.devspec.ai/action-items

An agent reserving and claiming explicitly requested work and recording evidence in DevSpec

The record keeps itself while the code gets written.

An Index of Capabilities

Plugged into where work happens

GitHub

OAuth repo connection, push webhooks for reindexing, and deployment webhooks that link deploys back to the work that shipped them.

Deployment-tracked commits

Agents and developers generate commit messages with [devspec:id] tags, so deployments connect to action items without anyone bookkeeping.

The same index everywhere

The semantic code search your sessions use is the same one your IDE agents query over MCP. One index, every surface.

Write-back protocol

An agent reserves an item before it claims it, its evidence and commits land on the item, and where the work goes next follows from that evidence. External agents are first-class citizens, not read-only guests.

Voice in and out

Deepgram-powered transcription and natural speech responses, built into sessions — no extra setup.

Convinced?

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

Start with $15of usage, on us · nothing charged until you buy