Roadmap Last reviewed November 2026

What we've shipped, what's next, and what we won't build.

13 things shipped since September. Checked against what actually merged, not what we hoped would.

What we're working on

Agencies and multi-team organisations

Several client teams or squads under one parent: one bill, each team's own numbers, and all of them side by side.
Who each repository rests on

People against repositories, shaded by their share of the work, so the repository only one person can touch stands out.
Set up by your coding agent

Tell your agent "set me up with Coderbuds" and it connects, then tells you the one step that is left.

A change failure rate that counts real failures

Reverts, fix-forward pull requests, main going red after a merge, and changes that never reached production, all read from git. No agent or production access needed.
Checks your coding agent writes

When it opens a pull request, the agent records what should be true once the change is live (that it ran, changed something, and did not do too much), then checks it in a later session.
An AI review that counts

Coderbuds' own review, or the AI reviewer you already use, settles a change your policy says an AI review is enough for, so it no longer waits on a person.
The fast lane

A change that fits goes to the reviewer best placed to approve it and merges fast. Standards are only worth keeping if keeping them is quicker.
Weekly summary by email

The weekly summary your Slack channel already gets, for the people who do not live in Slack.

A review policy that can hold a merge

The review policy check made a required check on GitHub, so a change merges once it has the review it needs and not before.
Receipts, published

What the fit check changed across every team using it, in size, merge time and how often the advice was taken. Each team already sees its own.
Teaching the org map

Facts an agent cannot observe, recorded with who said them and when they stop being true — sparingly, so it never becomes another wiki.
Earned autonomy

Agents acting without asking, only on the actions a team has already been approving by hand.

1 shipped

How much review each change needs

Choose a review policy and every pull request says, in a check beside CI, whether it needs no review, an AI review, one review or a careful one, and why.

5 shipped

Findings you can settle in one glance

Each Inbox finding says why it was raised and what it rests on, and can be answered from the row.
A Claude Code plugin

One install brings the size check before a pull request and the AI usage report after it into Claude Code, in every repository. Codex, Cursor and any other MCP client connect to the same server.
History imports that finish

Private repositories and restricted organisations import end to end, paying teams get their whole six months, and a rate limit is waited out instead of dropping pull requests.
Production, picked from your deploys

Coderbuds names your production environment from the deploys each repository reports, asks you to confirm it, and no longer rates a DORA metric it could not measure.
One price, every feature

$29 per active developer a month, published, with the bill for your team worked out on the card. No tiers and no sales call.

Show earlier months

7 shipped

The delivery bottleneck, named

The one stage holding delivery up — from backlog to production — who is carrying it, and the one change to try first.
One Inbox for every finding

Findings live in one place, your rating decides how long one stays away, and the Slack digest goes once a day, only when there is something to act on.
Activity: a changelog of what shipped

A readable feed of what each period shipped, with workstreams inferred from the pull requests and your epics laid out as rows.
AI attribution an agent can declare

A coding agent's own report decides a pull request's AI verdict, each MCP call records which agent made it, and we published how our detector grades per tool.
Agent follow-through, with receipts

Every agent action shows its stage and what came of it, and a recommendations-only posture lets a team take the advice without the actions.
Setup is one install

Installing the GitHub App is the whole of setup, and adding repositories is a slide-out where you can tick a whole organisation.
Dark mode, and a profile for everyone

The dashboard in dark mode, a rebuilt profile showing where each person's work went, and periods that reach a quarter or a year.

6 shipped

Coderbuds Standards and the pre-flight fit check

Coding agents call assess-change-fit before opening a pull request and get a verdict against how your team actually ships, calibrated on your own merged work.
The org map for coding agents

get-team-context gives an agent the live picture of how your repositories fit together, from manifests, co-change history and the links you teach it.
MCP: one-click connect, and tools worth calling

Connect Claude Code, Cursor or Codex over OAuth in one click, with tools for standards, pull requests, deploys, DORA and your own status.
Deploy tracking, and who actually deployed

Deploys credited to the person who triggered them, an inbound webhook for platforms that send no event, and deploys started from Slack or an agent.
The dashboard rebuilt around the loop

Nine tabs cut to four, an Overview of Today, Delivery and Health, and empty sections that say why they are empty.
Reports and the partner API

Board, leadership and engineering reports exported as PDF, Markdown or into Slack, and a REST endpoint for building your own dashboard.

1 shipped

The delivery loop, in Slack

A ready-to-deploy queue with a Deploy now button, /review on any pull request, and one weekly pulse in place of five summaries.

1 shipped

Agents, insights and integrations

The coaching agents, insights across the SPACE dimensions, DORA metrics, and connections to GitHub, Bitbucket, Jira, Linear, Slack and Microsoft Teams.

A roadmap is only useful if it rules things out.

A settings page for every threshold

The size standard calibrates on your own merged work and proposes moving when your overrides disagree. One switch, not a settings farm.
AI code review as the headline

Reviewing the diff is a crowded, well-served problem, and it is not where work gets stuck. Review stays a feature; our effort goes into landing the change.
An autonomy dial per agent

One team-level posture governs every agent, so there is a single answer to what agents may do here.

Tell us what you'd want next. It goes straight to the people building it.