Your coding agent writes the plan, the spec or the report into a Mirrorly doc instead of a file on your laptop. Share the link: teammates edit it, leave comments and tag their own agents, all in the same doc, live.
Mirrorly is in beta: join the waitlist and we’ll e-mail you when your access is ready. Already in? Sign in
A new team goes from sign-up to its first shared doc in five minutes.
Invite e-mails may land in spam for some partners.
Fallback: send invites from a warmed-up subdomain, and ask each partner to allowlist it.
Set it up once on the machine where your agent runs. Everyone else needs only a browser.
Add Mirrorly to your agent as an MCP server, and it signs in once, in your browser. In every doc it shows up as yours: Nika’s Claude Code, Nika’s Codex.
When it would write a plan to a file or paste a long answer into chat, it writes a Mirrorly doc and gives you the link. The doc is the only copy, so nothing goes stale.
Teammates sign in and work in the same doc: typing, commenting, tagging people and agents. Their agents can join too.
Today the plan your agent wrote sits in a terminal, in a file on one laptop, or in a message someone pasted into Slack. In Mirrorly it lands in the doc your team is already in, and every change says who made it.
Agents work in the doc the way people do: a caret where they’re editing, edits that land live, comments under their own name. Each session carries its repo and branch, so you can tag exactly the one you mean.
Rewrite the first-run screens around one action.
Invite e-mails may land in spam for some partners.
Every passage someone changed since you last read it is marked. Open its history for who changed it and why, the before and after, and restore any earlier version.
Beta opens to ten design partners on 10 November, a week after the first-run rewrite ships.
Weekly check-ins on Fridays, with the numbers in this doc.
Nika runs the rollout. Ana owns the invite e-mails and the partner list.
Moved the beta back a week
Why: the first-run rewrite slipped
Beta opens to ten design partners on 3 November 10 November, a week after the first-run rewrite ships.
Type @ in a comment to bring in a teammate or an agent. A teammate sees it beside the passage. An agent finds the request the next time its person asks “anything for me?”, and acts only once that person says yes.
Pricing: $12 a seat, billed monthly.
Waiting for Nika’s Claude Code
The repo says $14 a seat since the last release. I updated the line and linked the commit.
Thanks. @Ana good to share?
Ana is tagged in this thread
Agents can put a live chart, a dashboard or a clickable prototype right in the page. It runs sealed off from the rest of Mirrorly, and everyone edits the words around it as usual.
Retention held above 60% while sign-ups more than doubled.
Next: try the shorter first run with the next ten teams.
An agent built into every doc. It reads along while people and their agents write, and speaks up only when something needs everyone’s attention.
When people or their agents have edited and paused, Mirrorly takes a look. It works for the doc, not for one person.
Two sections that disagree, a summary an agent’s edit left stale, a question nobody answered. Mirrorly says so in the margin, beside the passage, for everyone to see.
A typo, or a list item typed as a paragraph, it fixes by itself, and the doc’s history shows the change, so anyone can restore what was there. Anything bigger becomes a comment, never a rewrite.
Tag @Mirrorly in a comment to review a section, find the gaps or draft a summary. It acts with the role of whoever asked: if you can only comment, so can it.
“Leave §Draft alone”, “focus on §Pricing”, “stop”, and later “go on”. Mirrorly keeps to what you said until someone changes it.
You stay in charge: everything Mirrorly changes is in the doc’s history and can be restored, you can resolve or delete anything it leaves, and it does only so much in each doc a day.
Agents work in your docs with the access you give them, and every change they make is yours to keep or roll back.
Roles work the same for people and agents: view, comment or edit. An agent never has more access than its person, and disconnecting it in Settings turns off every key it holds.
Changing or removing someone’s access, and deleting a doc for good. No agent can do either. An agent can share only up to its person’s role, and what it moves to Trash can be restored.
Your agents run on your machine, with your files. A teammate can ask your agent for something, and it acts only after you confirm.
Each edit says who made it, when and why. Restore any passage to an earlier version from the doc’s history, and an agent can undo its own edit.
Edits merge live, like in any shared doc. If the text an agent read has changed since, its edit is refused, and it reads again before it tries.
Install once, on the machine where your agent runs. Teammates who read, comment or edit just open the link and sign in.
Before your team writes in a doc, know who can open it and what the AI sees. Here’s how access, sharing and AI processing work in Mirrorly.
Only you, until you share it. A share link is for one doc or project and one role, people sign in to open it, and you can reset or turn off the link any time.
When Mirrorly responds in a doc, text from that doc, and docs linked from it that the person asking can open, may be sent to its AI provider. Your own agents, like Claude Code or Codex, use your own accounts with those companies.
Notes about each project, kept to what everyone in the doc can open. You can read, correct or delete every one on the project’s Memory page.
Docs, comments and images are encrypted in transit and at rest.
No. Mirrorly doesn’t sell your docs or use them to target ads.
16 answers
Join the Mirrorly waitlist, and we’ll e-mail you when your access is ready. Your next plan lands in a doc your team can open.
@Nika’s Claude Code can you check this against the pricing page in the repo?