Feedback board + AI product team, in one place

Customer votes in. Reviewed pull requests out.

Kindling collects ideas and votes like a classic feedback board — then goes further. Approve a request and Otto drafts the PRD, a coding agent builds it inside your own GitHub Actions, a second agent reviews the diff, and every voter hears the moment it ships. You stay in control: humans always merge.

Brainstorm with AIExplore public boards →
K
A
D
L
N
built for teams shipping in the open — with agents that answer to them
From idea to merged PR

One request, five steps, zero handoffs lost.

Every stage is visible on the board, so requesters, PMs and engineers watch the same thing move.

1

Collect & vote

Customers post ideas and upvote. Duplicates merge so the signal stays clean.

2

Approve

Mark a request approved. Otto drafts a PRD with testable acceptance criteria.

3

Build with an agent

One click dispatches a coding agent into your own GitHub Actions runner.

4

Verify & review

Tests run, quality gates check the diff, and a second agent reviews the PR.

5

Merge & close the loop

A human merges. Voters get a release note that links back to their request.

Why Kindling

Most feedback boards stop at “planned.” Kindling keeps going.

Votes are only useful if they turn into shipped software. Kindling wires the board directly to your repo, so the distance between “most requested” and “merged” is a PRD review and a PR review — not a quarter.

  • A feedback boardKindling
  • Collect votes, export a CSVCollect votes, generate a PRD, ship a PR
  • Status updates typed by handStatus, build timeline and release notes update themselves
  • Roadmap decided in a spreadsheetWeekly ranked build queue: votes × effort × risk, with rationale
  • Tech debt invisible to customersAgents audit the repo and propose fixes you approve in one click
What you get

A product team’s worth of help, on your terms.

Otto, your AI PM

Answers questions about the board, drafts PRDs from requests, merges duplicates, and posts plain-English updates on every request as builds progress.

PRDs with a testability gate

Every approved request gets a PRD. Acceptance criteria are linted for testability before the agent sees them — vague criteria get inline rewrite suggestions.

Forge: agents that open PRs

Claude Code, Codex, Gemini CLI or our built-in tool-calling loop. Runs in your GitHub Actions with your API key. Optional plan checkpoint before any code is written.

A quality loop, not a coin flip

Deterministic gates (tests pass, protected paths untouched, diff size), an LLM judge scoring the PRD criteria, a stuck detector, and automatic escalation to a stronger model.

Second-agent review

A separate Reviewer agent reads the diff against the PRD and leaves PR comments before a human ever looks. Its verdict shows on the request.

Close the loop automatically

When the PR merges, Kindling writes a release note from the PRD and diff, emails every voter, and links the PR, request and vote count together for traceability.

The agent collective

Four specialists that keep your backlog honest.

Beyond building features, Kindling runs a small crew of agents on a schedule. They never act silently — every finding lands as a proposal you accept, reject or ignore.

Read the guides
Scoutweekly

Audits the repo: outdated dependencies, flaky or missing tests, hot files, code smells. Findings become tech-debt proposals.

Reviewerafter every build

Reviews the agent’s diff against the PRD and comments on the PR before a human spends time on it.

Streamlinerafter every merge

Spots churn-heavy files worth refactoring and flags stale, obsolete or duplicate requests on the board.

Prioritizerweekly

Ranks the build queue by votes × effort × risk and explains each pick in a sentence.

You choose the speed

Turn autonomy up one notch at a time.

Each board has an autonomy level. Start at zero and move up as the agents earn trust. No level lets an agent merge to your default branch.

LEVEL 0

Assist

Otto drafts PRDs and answers questions. Nothing is dispatched without a click.

LEVEL 1

Build on approval

Approving a request scopes it and lets a PM start a build with one click.

LEVEL 2

Propose

Scout, Streamliner and Prioritizer publish proposals to the board for PMs to accept.

LEVEL 3

Self-serve tech debt

Agents build their own small tech-debt proposals inside a weekly budget and diff cap. PRs still wait for you.

Bring your own model keys — Anthropic, OpenAI, Gemini, Mistral or any OpenAI-compatible endpoint. Encrypted at rest, brokered to the runner per job.

Policies as code: protected paths, no-migration and no-dependency rules, allowed commands, required reviewers, PR labels.

Per-job cost caps and a weekly budget. A single kill switch disables every autonomous action on the board.

Every autonomous decision is written to an audit log. Agents open PRs; people merge them.

Explore

Public boards

See how other teams build in the open — watch requests move from vote to “building now” to shipped.

Loading boards…

Ready to ship what your users actually voted for?

Create a board in five minutes. Connect GitHub and a model key when you’re ready to let the agents in — no credit card, no setup call.