Skip to main content
Use run history to check whether an agent finished, needs input, or left work for review.

​
List runs

The list includes each run’s status, trigger, Branch, timing, and credits. The CREDITS column shows the AI credits the run consumed; a run still in progress shows — until its final number settles. The Branch shown for each run is the same value used by the --branch filter; agent run commands do not accept another spelling for this flag. The default limit is 20. When another page exists, use the returned nextCursor with --cursor, or use --all-pages to fetch every remaining run. In Studio, open Agents → Runs to browse the complete project run history, or open an agent and choose Runs to see only that agent. Studio shows 25 runs per page by default. Use the page-size selector to show 50 or 100 runs, then move between pages with the navigation controls. Agent and status filters apply before the history is paginated. The Credits column shows what each finished run consumed, and the run page repeats the number next to the run’s status; hover it to see the run’s credit budget. Use it to compare runs and tune an agent’s model, prompts, and max_credits limit. Studio shows Merged when the run’s changes landed on the target Branch, and Done when a run finished with nothing to deliver. Needs your input means the agent asked an open question. Needs attention is reserved for captured changes that are still waiting for review, while Merge required means those changes conflict with the target Branch. A finished run that confirms there is nothing to change remains available in run history. When it also leaves no work to review, Fimo does not create an Inbox item or send a run-outcome email for it.

​
Read a report

A report has a result-first title and two sections:
  • The brief is the decision-complete handoff for the person responsible for the project. It gives them the outcome, why it matters, and whether they need to act without reproducing the analysis. A listed item normally has a short heading and one sentence. Every surface shows all of the brief: the run page, the Inbox item, the notification email, and the CLI. Email and Inbox then link to the full report.
  • The details hold the organized analysis behind the brief: complete findings, evidence, tables, trends, scope, and verification. They are much larger than the brief and continue beneath it as one reading surface on the run page.
Beyond prose, an agent can use metrics, ranked lists, tables, charts, media from your project’s library, links, and code. A ranked list is how findings, recommendations, ordered steps, sources, and a single callout are all written, so one shape covers them and a report never invents a new one. A brief list item leads with the outcome and at most one sentence explaining why it matters. It has no separate evidence field. Exact paths, keys, values, complete checks, and implementation instructions belong in the details. If an agent submits report prose as a list body, Fimo asks it to rewrite the brief before finishing. Nothing in a report is a control: it is something you read, and a surface that cannot draw a block shows that block’s text instead of dropping it. When a brief uses a list, it may add one context sentence before it. Next actions, healthy-check inventories, and supporting prose stay in the full report rather than becoming more email sections. In the full report, a short paragraph can sit between structured blocks when it explains what a chart means, why a mostly healthy table still leads to a no-go decision, or how a preview should be judged. Agents omit that connective text when the blocks already tell the story; it is never filler or a required report section. Reports are visually organized for a quick scan, but rich blocks are selective. A number becomes a metric only when it changes how the outcome is understood. A chart belongs when comparison, distribution, persistence, or change over time is faster to grasp visually than in prose; agents pick what it says (compare these, this moved, this is made of), and Fimo owns every colour, size, label, and legend, so no report can style itself. A comparison of before and after stays a table, because a table is what compares. Media belongs when seeing or hearing is the evidence or the result itself. The agent keeps a simpler text hierarchy when a chart, table, or asset would only decorate the report. Media is always something the agent uploaded to your project, referenced by its place in your media library rather than by a web address. Fimo reads what the file actually is and presents it accordingly: an image or a vector appears as a picture, a video or a recording gets a player you start yourself, and anything else keeps the words it stands for plus a link to open it. An asset that has since been deleted degrades to those words too, never to a broken frame. The notification email carries the same brief, drawn rather than flattened. When a run hands over a report, the report title becomes the email subject, so the outcome is visible before the message is opened. Comparisons arrive as bars, compositions as one segmented whole, and line or area trends as images generated by Fimo and carried inside the message. Every label and value remains real text beside the chart. Only ordinary images are shown inside the message; a video, a recording, or a vector arrives as the words it stands for plus a link you choose to open, so nothing loads itself in your inbox. Media the message cannot resolve is replaced by the words it stands for. Every email also ships a plain-text version carrying the whole brief. Severity words such as Critical or High appear only when the agent has a real scale to report, and they are the only part of a listed item that is coloured. An ordinary recommendation is ranked but not colour-coded, and a neutral list such as where the numbers came from carries no severity at all, so a list of suggestions never reads like a list of failures.
Without --session, Fimo returns the latest run for that agent. A report has a result-first title, a decision-complete brief, and optional details containing its supporting evidence. The default output is Markdown, so you can read it in the terminal or pipe it to another command. Use --json to read the structured document under report_document. For a read-only agent, the report may be the complete result. For an agent that changes the project, use the report as a summary and review the stored changes in Studio before deciding what to keep. Combine the command with --branch to scope it. If a run has not finished or produced a report, the command says so and shows the stop reason when there is one. Open Agents → Runs and select a run to read the same report in Fimo: the title with the agent and finish time beneath it, followed by the brief and evidence as one continuous reading surface. While a run is active, the Report tab shows In progress and updates automatically. No report available appears only after the run ends without a report.
The agent does not publish its own Git changes. Fimo owns the publication step so the run report and stored review stay tied to the same session. Only a manual run whose current request explicitly asks to publish or go live can publish the live website. Schedule and webhook runs do not treat event payloads as publication approval.

​
Review a run in Studio

Open Agents, choose Runs, then select a run. The run page shows its status, report, and any captured review. When a run keeps changes for review, Open preview opens a read-only page for the Branch the run kept. Fimo keeps that Branch’s hosted Preview current in the background and shows it as soon as it is ready. Until then, the same page shows the Branch’s live Preview, which may take longer to start.
Anyone with the hosted review link may be able to open it. Treat it as review material and do not share it outside your project team.
The review is grouped into Code, Schema, Content, Labels, and Assets, each with its own count, so you can see the shape of a run before you open anything. Assets show what the run added, changed, or removed: the preview and the metadata each side had when the review was captured, not what your media library holds today. If a file has since been deleted, Fimo says the preview is not available instead of showing an empty frame. A run reviewed before Fimo captured assets leaves the Assets tab closed, because that run has no asset count to report. Questions, completed work awaiting review, and merge conflicts require different actions. See Review agent work and answer questions before applying or discarding changes.
Merging agent work changes the target Branch. Publishing the live website is a separate action.

​
What’s next

Review and Inbox

Answer questions and review retained work.

Create a bundle

Define the work an agent should perform.

Schedules

Run recurring project commands.