Product
Everything Backlog.cloud does, in the order it happens
A meeting goes in one end. Jira tickets, a requirements document, a decision log, and a process map come out the other. This page walks the whole pipeline, with the real numbers and the real limits.
The bot is booked before you think about it
Two ways in. Connect a calendar once and forget it exists, or paste a link when a call comes up. Either way, Backlog.cloud knows what kind of meeting it is walking into before it joins.
Calendar auto-join. Team plan
Connect Google Calendar or Outlook once. The connection is read only: we never send invites, decline meetings, or touch your events. From then on the bot turns up to any event with a Zoom, Meet, or Teams link that you are attending.
The ceremony type is inferred from the event title. When confidence drops below 0.7 it falls back to the project default rather than guessing. Need the bot out of one sensitive call? Skip that single event from the Meetings page and leave the connection on.
Paste a link. Every tier
No calendar connection needed. Paste a Zoom, Meet, or Teams URL and the bot joins in 30 to 60 seconds. It arrives under the name Backlog.cloud and waits to be admitted if the host runs a waiting room.
It listens only. It never speaks, never shares video, never presents. If it cannot get in, the meeting sits in Pending so you can cancel and redispatch with a corrected URL.
26 ceremony types, each with its own rules
Sprint planning, standup, retro, refinement, review, discovery, user interview, steering committee, incident review, architecture review, hiring debrief, 1:1, and 14 more. The type you pick decides what the extractor is allowed to produce, which cheat sheet you get, and which drift warnings arm during the call. A standup yields actions and follow-ups, not epics.
Pick the 1:1 type on the Team plan and the meeting locks to the two people on it. Nobody else in the workspace can see the transcript, the briefing, or the artefacts.
A one-pager before you walk in
The cheat sheet is on screen before the call starts: the agenda, the timebox, who should be in the room, the action items still open from last time, and the RAID flags on this project. It freezes on the meeting, so the prep snapshot is still there six months later.
You run the meeting. It takes the notes and watches the clock.
The bot sits in the call and streams the transcript into your dashboard in real time. You do not have to look at it. Most people keep the sidebar open anyway, because it earns its place.
The agenda ticks itself off
As the team covers each step from the cheat sheet, the sidebar marks it done. What is still uncovered stays visible, so the last five minutes go to the topic you actually skipped, not a vague feeling that something got missed.
Drift warnings, not etiquette tips
Standup at minute 8 with no blockers mentioned. One person holding the floor for 12 minutes. A planning session heading for the end with zero decisions on record. Each one triggers a browser notification, even when the tab is in the background.
You stay in charge
Pause or mute coaching mid-call when the conversation takes a turn that does not need a robot's opinion. Cancel a dispatched bot at any point. Bot time is metered from join to leave, to the second, against your plan's hours.
Minutes later, the whole meeting is structured work
The extractor produces up to 11 artefact types from a single call: user stories, bugs, tasks, epics, spikes, a requirements document, RAID log entries, a decision log, action items, meeting minutes, and a BPMN 2.0 process model. Which of those it is allowed to produce depends on the ceremony type, so a standup does not spawn epics.
Every artefact shows its source
Each item carries the verbatim quote it came from and a confidence rating of high, medium, or low. When the model had to infer something, the open question is attached to the artefact instead of buried inside it.
A second pass checks the first
A verifier pass rereads the transcript after extraction. Anything whose quote is not actually in the conversation gets dropped, and contradictions between artefacts get flagged before you push a single ticket.
Nothing is locked in
Edit any field, regenerate a single item, or re-tag the whole meeting under a different ceremony type and re-extract from the same transcript. A re-tag does not spend another generation.
You also get a briefing on top: the TL/DR, a forward timeline of every date someone committed to, the open loops nobody closed, and what to probe next time. Below is the output all of that grounding buys, from a real demo meeting.
Artefacts
This is what comes out. Not summaries. Not drafts. Production-ready artefacts, grounded in what your team actually said.
Every meeting opens with a TL/DR, a 4 to 6 week forward timeline of what is now committed, and the gaps your team did not close on the call.
Then it leaves. Six trackers, plus every format people ask for.
Artefacts are not meant to live in Backlog.cloud. Push them into the tracker your team already uses, with field mapping configured per project. The first push creates the issue, the next push updates it instead of duplicating.
Jira
Descriptions land as native Atlassian Document Format blocks, so headings, lists, and tables render properly. Bulk pushes back off automatically against the Jira Cloud rate limit of 65,000 points per hour.
Linear
Issues arrive with priority, project assignment, and prefixed labels, so pushed labels never collide with the ones your team already uses.
GitHub Issues
Acceptance criteria and reproduction steps become task-list checklists in the issue body, so the GitHub progress bar lights up.
Azure DevOps
Work items land in your configured area path with HTML descriptions, because Azure DevOps does not accept Markdown.
Notion
Each item becomes a database row with Status, Type, Priority, and Story Points mapped to your existing properties. Rows, not pasted text.
ClickUp
A ready-to-import CSV with priority translated into ClickUp vocabulary and tasks landing in the list and status you set as default.
And the exports that are not trackers
Files your stakeholders actually open
A multi-sheet XLSX workbook with one sheet per artefact type. A per-platform CSV zip. Word documents for the minutes, the decision log, the action log, and the RAID log. A .bpmn file plus an .svg of the diagram for process models.
Email the summary
Send the meeting summary to up to 10 recipients with a personal note on top. Useful for the sponsor who will never log into anything.
Secure share links
Share a link that expires after 30 days. The recipient signs in to view, and no meeting content travels in the email itself.
No bot on the call? Nothing is lost.
The live bot is the front door, not the only door. Anything that captured the conversation can go in after the fact, and every input runs through the same extraction, grounding, and verification as a live meeting. Same 11 artefact types, same exports.
Text
Paste a transcript, meeting notes, or a one-paragraph brief straight into the editor. The fastest path: a few seconds per artefact.
Images
Up to 8 whiteboard photos, Miro exports, or screenshots per generation. JPEG or PNG up to 10 MB each, HEIC converts on upload.
Audio
MP3, M4A, WAV, or OGG up to 200 MB. The file is deleted 30 days after upload. The transcript and the artefacts stay.
Video
MP4, MOV, or WebM up to 500 MB. We pull the audio track and delete the file after 14 days. The transcript and the artefacts stay.
Trust & Security
Encrypted in transit (TLS) and at rest (AES-256).
Your meetings aren't used to train AI models.
Built on SOC 2 Type 2 certified infrastructure.
Operated by Alconbury Tech Ltd, a UK company. UK and EU GDPR apply to how we handle your data.
Run one real meeting through it
The Free tier is the full product with hard caps, not a demo. 3 generations, 1 project, and 1 bot dispatch, for life. All 11 artefact types, every input, every integration. No card. Enough to put your next planning session through the pipeline and judge the output yourself.