OpenUI Live Replies

Turns agent replies into dashboards, reports, plans and forms you can read at a glance and click, right in the chat. Agents write OpenUI Lang, and the plugin draws it with OpenUI's React component library plus about 70 components of its own: gauges, heatmaps, timelines, roadmaps, kanban boards, test and coverage reports, diffs and maps. Plans, decisions, choices, quizzes and forms send your answer back into the thread as your next message. Components for BB show the thread's changed files, commits, pull request and other threads live. Adds about 3,000 tokens of instructions to sessions whose replies you read, and Map and ImageSearch fetch from OpenStreetMap, Photon, Wikipedia and Wikimedia Commons unless you turn web lookups off.

Overview

View source

What you get

Agents put OpenUI Lang programs in their replies (```openui-lang blocks). BB draws them inline with OpenUI's React renderer and its component library: cards, tables, charts, tabs and forms.
About 70 more components from this plugin: gauges, sparklines, heatmaps and funnels; timelines, calendars, roadmaps, kanban boards and estimates; trees, JSON explorers and diffs; test, lint, coverage, CI and benchmark reports; maps, credited photos and small sandboxed apps.
Plans, decisions, choices, quizzes, follow-up suggestions, buttons and forms send the user's answer back into the thread as their next message.
Components for BB itself show the thread's changed files, commits and pull request, and the live status of other threads and environments.
check_openui, openui_examples and openui_components tools let an agent check a block, see working examples, and fetch a component's signature before it uses it.

Requirements and costs

BB 0.49 or later (uses the experimental code-fence renderer slot).
About 3,000 tokens of instructions in each session a person reads. Worker threads get none. Signatures for most components load only when an agent asks for them.
No OpenUI Gateway and no API key. Gateway's automatic error correction is not used, so a block with no renderable root shows as code.
Web lookups, on by default: Map sends place names to Photon and loads tiles from OpenStreetMap, Map also queries Wikipedia, and both Map and ImageSearch query Wikimedia Commons. These services see your IP address and the request. Turn off the plugin's Web lookups setting and nothing is fetched.
Blocks also load the image URLs the agent writes (OpenUI's prompt suggests picsum.photos), and source/citation components load site icons from Google's favicon service. These see your IP address and the domains shown.
Form input is not saved: it resets when the reply scrolls out of view or the thread reloads.
Running it together with genui gives agents two UI languages and both prompts. Turn one off.

More from Muhammad Ahsan

GitHub PR DashboardSee why each pull request is stuck, and get it unstuck, without leaving bb. My PRs lists your pull requests, review requests for you and your teams, the ones you reviewed, and assigned ones by repository, each with its +/- lines and led by what blocks it: failing CI down to the job, changes requested, unresolved comments, conflicts or a stack. Reviewed turns amber when the author pushes after your review. Open one for its reviewers and their verdicts, labels, and tabs: Overview (the description, then comments and reviews), Checks with error annotations, Files with syntax-highlighted diffs and review threads under their lines, Commits with the ones since your review marked, and Comments. Edit the title and description, comment on a line or range, reply, delete your own comments, submit a review, mark your draft ready, and close or reopen, each only when you click. A menu links the PR to a bb thread, asks about it, explains it, or sends every finding to an agent, starting in the PR's project; Fix with agent does it for one comment or failed job. Jira keys open in bb's My Jira plugin. Search, hide or fold repositories, move by keyboard, switch gh accounts, and get alerts when CI fails or someone comments. Uses the GitHub CLI login or a token; github.com only.New Usage ExplorerSee how much of your Claude plan is left and where your AI coding spend went, by session, day and model, across every agent CLI you use. Each limit window shows what is left, when it resets and where you will land at your current pace, with a plain verdict on whether there is room for a big task. Sessions carry their titles and projects; open one for its models and its cost by token type, or click a model to filter the whole page. Reads Claude Code, Codex, Gemini CLI, OpenCode and the other CLIs ccusage supports from their local logs, and draws each provider in its own colour. Prices come from LiteLLM's public price list, so a newly released model is priced without a plugin update. Plan limits use the Claude Code sign-in already on your machine and need a Claude subscription.New Jira BoardSee every Jira ticket assigned to you in one bb tab, filtered and grouped the way you work, and move it along without opening Jira. Filter by status and type, search by key or words, and group by status, initiative and epic, or the boards that list each ticket. Open a ticket beside the list to read its lineage, description, subtasks and comments, then step to the next one with Previous and Next or the keyboard. Change a status in place from the moves your workflow allows. Needs Jira Cloud with Atlassian's Rovo MCP server allowed for your organization; you sign in with Atlassian in the browser and set up nothing else.New

More in Thread Content 89

Change what people see or do inside an open thread.

View all

Generative UITurns agent replies into charts, diagrams, maps, plans and forms you can use, right in the chat. Agents choose from 86 components on their own, with no setup: sliders and filters recompute in the page, and buttons, choices and forms put your answers in your message, which reaches the agent when you send it. Test reports, logs, lint errors and coverage files render as their own components instead of pasted text, and a thread's changes, commits, pull request and workers are read live from BB. Adds about 6,400 tokens of instructions to sessions whose replies you read, and looks up photos, places and map tiles on Wikimedia Commons, Wikipedia, Photon and OpenStreetMap unless you turn web lookups off. Relies on BB internals, so a BB update can show the blocks as code until the plugin catches up.

ComicBit894

Agent BoardFollow live BB agent activity, freshness, completion outcomes and monitoring health. Filter and group visible work by project, repository, provider or status. Optional Ollama Fleet and Redteam add worker visibility. Requires BB 0.40+; BB 0.46.0 recommended.

Aaron Phifer363

Agent TracesSee where your agents' time goes, from every thread down to each command, tool call, and subagent of a turn. A turn opens as a timeline that splits the agent's time into waiting on you, tools, subagents, thinking, writing, and model time, and it grows live while the agent works. Select a call to see its exit code, output, and arguments, or jump to the reply that followed it in the chat. A Traces page in the sidebar compares your recent threads by active time, calls, and errors, and two turns can be compared phase by phase. Turns export as OpenTelemetry or an HTML snapshot, and the bb agent-traces command gives agents the same answers. Traces come from the thread history on your bb server; an OpenTelemetry collector is optional.

Nicolas Charpentier277