The web companion to the CLI. Your team's shared bucket, trails, context tree, and datasets become browsable. This is the real prototype, running in your browser. Drive the whole thing below, or scroll for a page-by-page tour of every view.
interactive prototype
The live Hub preview is built for desktop. Come back on a larger screen to drive it — for now, here's what each view does.
Get early access for you and your team, a 15-minute call with the founder to get you set up, plus unlimited trace intelligence credits while the Hub is in preview.
A live look at what we're building into the Hub. Every panel below is the real view, booted on its own, not a screenshot.
A year of agent runs at a glance: a contribution heatmap, token spend by project, dataset sync state, and a trace ledger you can group by day, project, or dataset.
See what's happening in a codebase right now: branch, open and total traces, last push, the working tree's modified and added files, and the trace in flight.
Replay any run as a conversation: prompts, responses, reasoning, and every tool call inline, with git commits surfaced as checkpoints in the stream.
Switch any trace to its trail: a lane-by-lane timeline of the whole run — user turns, planning, reads, edits, and git work — with context-waste and run health flagged inline.
Every PR carries the agent trail that built it: which commits are traced, which stated intents the code actually satisfies, and the steps, cost, and wall time behind the diff — reviewed commit by commit.
Ask anything in plain english. Spotlight reads bodies, reasoning, and tool calls with an LLM grader, then returns ranked traces with scores, snippets, and fingerprints.
Group your traces by what the agent was actually doing: debug investigations, refactors, tool thrash, spikes. Edit an intent, regrade the window, and pull up every matching trace.
Write the SOPs your agents should follow. An LLM judge grades every run against each rule statement, counts violations, and links you straight to the offending traces.
Seal a failing agent session into a replayable capsule, attach it to a repo issue, and re-pose the intent against current main. The verdict moves from fails @sha to passes @main on its own.
Write if/then rules over failure rate, context overflow, tool thrash, cost, and SOP violations. Each fires on a cadence and pipes the offending trace IDs straight to Slack.
When intent spikes or SOP violations cross your thresholds, the agent drafts a fix and opens a PR against the connected repo. Auto-merge stays off, so you stay in the loop.
Your workflow drops new rows into an inbox. Promote the ones worth keeping to staging, skip the rest, then push the batch to a Hugging Face remote.
A workflow is the recipe that turns runs into dataset rows. Attach security and privacy tools to it: secret scanning, PII redaction, path anonymization, and your own redaction rules all run before a single row lands.
Get early access for you and your team, a 15-minute call with the founder to get you set up, plus unlimited trace intelligence credits while the Hub is in preview.