Human input on agent-made docs

Your AI agent wrote the doc.
Now get human feedback on it.

Share the specs, proposals, and reports it writes. Teammates and clients comment and answer your open questions in context — your agent gets it back as structured feedback before the work moves forward.

For the document, not the finished product — not live sites or shipped apps.

Start free
Founding pricing Unlimited reviewers No account to review
preview · checkout-spec.html
Spec · v2 · 4 threads
Checkout redesign

§2 Payments. Guest checkout must support Apple Pay at launch, alongside card and PayPal.

§3 Errors. Every failed payment shows an inline, recoverable error state.

claude code · markloop mcp
> apply the review on checkout-spec and revise it
markloop_pull checkout-spec
  ⎿ v2 · comments + answers · target · quote · intent
Edit checkout-spec.html ✓ §3 errors + 2 fixes
markloop_reply_comment #1 · noted: Apple Pay in v2
markloop_put_version resolved: 3
  ⎿ v3 published · all threads resolved
How it connects

Native MCP for Claude Code and Codex — push and pull without leaving the terminal. Any other tool? The same feedback exports as plain files.

The old loop

Your agent writes faster
than your team can review.

Your agent hands you a polished spec, a research report, a design doc — a clean HTML artifact. To get feedback you paste it into a Google Doc (and the formatting dies), screenshot it, or drop vague notes in Slack. Then you hand-translate every comment back into a prompt.

01

Formatting dies in transit.

Paste an HTML artifact into a Doc and the layout, code blocks and structure fall apart.
02

Feedback scatters.

Slack, email, screenshots — nothing anchored to the actual line it's about.
03

You become the middleman.

You hand-translate every comment back into a prompt for your agent.
See it in action

The whole loop, in two minutes.

Push a doc, collect anchored feedback, pull it back into your agent — watch one full pass end to end.

Markloop demo — feedback loop walkthrough
The Markloop loop

One loop, from feedback to the next version.

1
2
3
4
ON LOOP
1

Push

Your agent renders an HTML doc and pushes it to Markloop over MCP — one tool call, no manual upload. Reviewers get a link; nothing to install.

2

Collect

Anyone you invite comments on the exact heading, sentence or element — pinned right where it belongs. No account, no agent needed on their side.

3

Pull

Your agent pulls every open comment back over MCP — selector, quote, intent, version — as structured context it can act on in one pass. No copy-paste.

4

Apply & repeat

Your agent applies the changes locally and pushes the new version. Markloop checks off what's addressed and keeps every thread.

What's inside

Precise feedback in.
Agent-ready changes out.

Anchored comments

Pin a section or a single sentence. Comments stay anchored to the exact spot and stay attached to the version they were made on. Built for dense specs and docs.

1
2

Agent-ready package

Feedback your agent can read. Every comment exports with its selector, quote and context — ready to apply in one pass.

## Comment 3 · OPEN · v2
Target: h2#payments
Quote: "Apple Pay"
Note: Confirm Apple Pay is
      in scope for v2.

Your work stays private

Reviewers see your rendered doc — never the raw file.

Versions that remember

Comments live on their version; new revisions check off what's done.

v1
v2✓ 3

Unlimited reviewers

Invite clients & teammates as guests — never charged per reviewer.

MK
JL
RW
+5

Threads & decisions

Discuss in place, resolve, and mark status.

Can we tighten the spacing here?
Done in v2 — closing this.
The package

The package is the whole point.

Other tools collect feedback. Markloop hands it to your agent. Every open comment comes back over MCP — or as plain structured files — with target selector, reviewer intent, surrounding context, and the version it belongs to, so the agent applies the whole batch in one pass. No copy-pasting comments into prompts. No guessing what "that section" meant.

Plain files. Any agent. Nothing runs on our servers — you apply changes locally and stay in control.

markloop-package/
markloop-package/
├─ index.html         ← the current version
└─ comments.md       ← the review, for an agent

## Comment 3 · OPEN · v2
Target:  h2#payments     (css)
Quote:   "Apple Pay"      (occ. 1)
Note:    Confirm Apple Pay is in
        scope for v2 — contradicts §3.
Where:   §2 Payments
Who it's for

For everyone whose agent writes things people need to weigh in on.

Get sign-off on the specs your agent drafts.

Share a PRD or product spec, collect anchored comments and answers to your open questions, and let your agent fold them into the next version. No more Doc-paste graveyard.

Review technical design docs in context.

RFCs, architecture docs, migration plans — comment on the exact line, answer open questions, and track what's addressed across revisions. The feedback comes back ready for your agent.

Share polished deliverables, collect client feedback.

Reports, proposals, audits — send a professional HTML artifact instead of a messy attachment. Clients comment and answer the questions you embed, all in one place, scoped to just their project.

Loop feedback on any doc your agent makes.

Research, strategy docs, plans, analyses — if your agent writes it as a doc, you can share it, gather comments and answers, and loop the feedback straight back.

Security

Share the artifact. Keep control.

Your work stays private

Reviewers see your rendered doc in the browser — never the source file.

You control access

People see only the projects you invite them to. Nothing is shared by default.

Scoped guests

Invited reviewers see a single project — never your team, other clients, or billing.

Pricing

Pricing that scales with your work.

Start with a 7-day free trial — cancel anytime. Unlimited reviewers on every plan; you're never charged for the people leaving feedback.

Solo
For one builder working with an agent.
$19/mo
founding price · billed monthly
Start free trial
$0.00 due today · cancel anytime
  • 1 creator seat
  • Unlimited projects, docs & versions
  • Unlimited reviewers (always free)
  • MCP for Claude Code & Codex
  • Founding price, locked in for early users
For teams
Team
For a team building together.
$49/mo
founding price · billed monthly
Start free trial
$0.00 due today · cancel anytime
  • Everything in Solo, plus:
  • Up to 5 creator seats
  • Shared workspace
  • Team roles & permissions
  • Remove Markloop branding
  • Priority support
Need something specific?
Bigger team, a custom setup, or special requirements — let's figure it out together.
Talk to us
How it compares

The only feedback tool built for agent-made documents.

MarkloopGoogle DocsNotionSlack + screenshots
Plugs into your AI agent — push & pull over MCP
Keeps your agent's HTML intact (layout, code, diagrams)
Anchored to element and exact text quotetext onlytext only
Export feedback as an AI-agent-ready package
Versions with "addressed" trackingpartialpartial
Reviewers can't download your source filen/an/a
Unlimited reviewersvaries

Google Docs and Notion are great for writing; Markloop is built to gather feedback on the docs your agent produces — comments, answers and all — and feed it straight back to the agent.

Questions

Everything else.

Do reviewers need an account?

No. Invite by email or share a link — reviewers are always free, with nothing to install.

Which AI agents does it work with?

Claude Code and Codex connect natively over MCP — your agent pushes and pulls without leaving the terminal. Any other tool still works: the feedback also exports as plain files, so bring your own.

Does Markloop change my work on its servers?

No. Your agent applies changes locally, on your machine. Nothing runs on our servers — you stay in control.

Is my work private?

Yes. Reviewers see your rendered doc, never the source file, and guests are scoped to a single project.

What can I upload?

HTML artifacts your agent produces — specs, docs, reports, plans, research. Anything that renders as a self-contained HTML file. Each upload becomes a version in a chain.

Does it work for live websites?

Markloop is built for self-contained HTML artifacts and documents — the kind your agent outputs — not live, multi-page web apps. If it renders from a single HTML file, it works.

Is there a free plan?

No perpetual free tier, but every plan starts with a 7-day free trial — cancel anytime before it ends and you won't be charged. Reviewers are always free.

What counts as a "project"?

A project is one client or workstream — it holds folders, files, versions and the reviewers you invite. Access is scoped per project.

Can I cancel anytime?

Yes. Plans are monthly or annual — cancel whenever, no lock-in.

Close the loop between your agent and the people who review it.

Your agent pushes, people comment and answer in context, your agent pulls the feedback and revises the next version — over MCP. Start your free trial today.

7-day free trial · Founding pricing · Reviewers always free