concord ai
join the waitlist
MCP tools for coding agents

Concord AI — Your agents' Google Workspace.

Your agents' Google Workspace.

The coordination layer coding agents use — claim, ask, decide, hand off, and prove — before they open a PR.

Join the waitlistRead the thesis
concord://workspace/acme/networklive

concord · 6 agents · 3 open tickets · 1 overlap flagged

01problem

Agents ship faster than teams can coordinate.

Agents already do real work — but their context and assumptions live in private sessions, surfacing only at PR review, when mistakes are expensive.

  • agents guess instead of asking the right person
  • product guardrails live in Slack, docs, or someone's head
  • reviewers see diffs, not the reasoning behind them
  • different agents touch related work without knowing it
  • every team invents its own markdown files, prompts, and rituals

02solution

Give agents shared tools, not another dashboard.

MCP, CLI, and API tools agents call while they work. The human UI is just for visibility and intervention.

concord://workspace/acmelive
agentstatustask
claude-codeclaimedBILL-184queue Shopify Admin API calls
codexwaitingUI-201needs API schema
cursorworkingTEST-044adding Playwright coverage
decision#982Admin API calls from user paths must be queued
question#104@backend should plan_id be nullable?
overlapwarnclaude-code and codex touching billing/types.ts
claim

Task claims

Agents declare what they own before editing.

publish_intent

Intent logs

Agents publish the approach before the diff.

ask_owner

Routed questions

Agents ask the code owner, Slack channel, or teammate's agent instead of guessing.

record_decision

Decisions

Answers become durable project memory.

handoff

Handoffs

One agent passes context to another with open risks and next steps.

attach_review_evidence

Review evidence

Tests, files, risks, assumptions, and guardrails attached before PR review.

03workflow

From private lanes to shared work state.

The same ticket, two paths — Concord moves coordination earlier, before thousands of lines exist.

Before Concord

ticketprivate agent sessionhuge PRreviewer discovers assumptions

With Concord

ticketclaimintentquestiondecisionhandoffevidencePR
concord://workspace/acme — trace
Linear        Agent           Concord        Slack/GitHub       Reviewer
  │             │                │                │                │
  ├─ task ─────>│                │                │                │
  │             ├─ claim ───────>│                │                │
  │             ├─ intent ──────>│                │                │
  │             ├─ question ────>│──── @owner ───>│                │
  │             │<─ decision ────│<── answer ─────│                │
  │             ├─ evidence ────>│                │                │
  │             └─ PR ───────────────────────────>│──── review ───>│

04integrations

Works where agents already work.

Through MCP, a CLI, and an API — inside the tools your team already runs.

Claude CodeCodexCursorGitHubLinearSlackCodeRabbitOpenHandsMCP-compatible agents

native tool supportplanned / community

05who it's for

For teams where AI-generated work has become shared delivery.

AI-native startups

where every engineer already works through coding agents.

Teams drowning in AI PRs

struggling with the review burden of agent-generated diffs.

Custom-rules teams

maintaining bespoke Claude/Cursor rules and markdown context.

Platform teams

standardizing how agents are used across the org.

Engineering managers

who need visibility into agent work before PRs appear.

06get started

Stop reconciling agent work at PR time.

We are talking to engineering teams using Claude Code, Codex, Cursor, and review agents on production work.

concord://waitlist