bluedoc

Agent Review Pages

bluedoc is an agent skill that turns what a coding agent writes, from plans and code reviews to architecture docs, UI designs, slide decks and store listings, into interactive pages. The agent writes one JSON file, and a local server renders it as a page you comment on, decide and approve. Your agent reads every answer and gets on with the rest. It runs on 127.0.0.1, with no accounts and no hosted service, and is open source under the MIT license.

bluedoc is an agent skill that turns what a coding agent writes, from plans and code reviews to architecture docs, UI designs, slide decks and store listings, into interactive pages. The agent writes one JSON file, and a local server renders it as a page you comment on, decide and approve. Your agent reads every answer and gets on with the rest. It runs on 127.0.0.1, with no accounts and no hosted service, and is open source under the MIT license.

Location

Cluj-Napoca

Oct 2026 - Present

Industry

AI Developer Tools

Challenge

Agents answer in walls of text and wait for a reply in prose. Reviewing a plan or a code review in a chat means quoting passages, restating options and hoping every decision gets picked up.

The goal was a page that asks the questions and keeps the answers, works with any agent, and stays fully local, without any HTML for the agent to keep in sync.

Results

• 8 doc types: docs, plans, reviews, designs, presentations, websites, store listings and other pages such as status reports and meeting notes.

• 6 agent plugins: Claude Code, Codex, pi, omp, Copilot CLI and Antigravity CLI, plus any other agent through the skills CLI.

• One JSON file per doc: the local server renders the page from it, so there is nothing to keep in sync.

• One loop from draft to approved: the agent writes, you mark up with pins, passages or drawings, you pick an option per question with the agent’s recommendation starred, the agent revises, and you approve.

• Reviews pinned to code: findings sit on their lines from git, with one decision per finding.

• Approval gate: each round is a revision you can compare, and the agent waits for approval before it writes code.

Stack

Stack

Stack

Conclusion

bluedoc replaces the back-and-forth of chat with a page built for decisions. Every doc an agent writes lands in one place, listed by type, status and revision, and every comment and choice goes straight back to the agent.

Released as v1.0, it stays local-first and open source, and installs in a couple of commands on every major coding agent.

bluedoc replaces the back-and-forth of chat with a page built for decisions. Every doc an agent writes lands in one place, listed by type, status and revision, and every comment and choice goes straight back to the agent.

Released as v1.0, it stays local-first and open source, and installs in a couple of commands on every major coding agent.