Solutions

Every way teams turn an HTML report into a conversation.

Comma is one product with eight ways into it. Pick the surface that matches the work in front of you — reviews, automation, or workspace. The render engine and the comment layer are the same underneath.

01 · Reviews

Reviews that happen on the document.

Anchored, in-context feedback for the artifact itself — not a screenshot, not a copy-paste.

AI engineers · researchers
01

Review AI reports as they regenerate.

Snapshot each eval run, anchor comments to the rows that changed, and let your agent post the next revision under the same link. Reviewers see deltas, not diffs.

What you get
  • Faithful render of any model output, charts intact
  • Anchored threads survive revision
  • Compare two snapshots side-by-side
Data leads · analysts
02

Stop pasting Looker screenshots into Slack.

Export the dashboard HTML, drop it in Comma, and stakeholders comment on the actual cells, not a JPEG. Threads stay anchored when you re-run the query.

What you get
  • Sandboxed iframe, scripts and tables intact
  • Cell-level pins on tables and charts
  • Re-upload to refresh — threads follow the content
PMs · ops · finance
03

Run quarterly reviews on the document itself.

Quarterly business reviews, launch readouts, post-mortems. Share one link, gather feedback in-line, resolve threads when shipped.

What you get
  • Roles by email or share link
  • @mention notifications
  • Resolve threads when work lands
02 · Automation

Reports that ship themselves.

Agents and schedules post the next revision while you sleep. You wake up to a refreshed thread.

Claude · Cursor · VS Code
04

Let your agent ship the report for you.

One line installs Comma as a remote MCP server. Your agent uploads HTML, leaves anchored comments, and replies to feedback under a scoped, revocable token.

What you get
  • Streamable HTTP MCP
  • Same surface as the REST API
  • Per-token rate limits and audit
Recurring reports
05

Reports that refresh themselves.

Wire a skill to a cron. Comma re-runs it on schedule, posts the fresh HTML as a new revision on the same report, and pings reviewers.

What you get
  • Monthly on Free, daily on Pro, hourly on Team
  • Bring-your-own Bedrock or +20% markup
  • Per-run cap so a runaway prompt can't drain you
Reusable generators
06

Package a report once. Run it forever.

A Skill is a versioned prompt + data source pair that always emits the same shape of HTML. Reusable across routines, agents, and one-off runs.

What you get
  • Version-controlled prompt and schema
  • Shared inside your team
  • Powers both routines and on-demand runs
03 · Workspace

Workspace controls when teams scale.

Sources, seats, and structure — the boring-but-load-bearing layer that makes the rest safe.

Source integrations
07

Bring your data without writing a fetcher.

Connect the warehouses, dashboards, and APIs your reports already read from. Skills pull from connectors so routines don't ship credentials in prompts.

What you get
  • Credentials kept out of prompts
  • Reusable across skills and reports
  • Scoped per workspace
Squads · org-level controls
08

Workspaces with the controls a team needs.

Folders, roles, audit log, DNS-verified domain, IP allowlist, and team API tokens. SSO and a custom domain on Enterprise.

What you get
  • Folders for structure, roles for access
  • Audit log and IP allowlist (Business)
  • SSO and custom domain (Enterprise)
One product, eight angles

Pick the surface. The link is the same.

Reviews, automation, and workspace are different doors into the same render-and- comment surface. Start with the one that matches your week.

Not sure which fits?

Most teams start with a single review surface — a dashboard, a weekly readout, or an eval — and grow into automation. Routines and agents click into place once the review habit is there.

Watch the 90-second demo