Pastel collects the feedback on the page.
Snag turns the bug into a pull request.

Pastel is a review canvas: point it at a live web page or a video, share the link, and anyone can leave in-line comments on top of it without making an account. Snag works the other way round. The snippet already lives on the site, so a report is filed from the screen the problem happened on, and it leaves the browser as a session replay with the console and the network beside it. Personal data is masked in the browser before anything uploads — inputs are never recorded. That becomes a ticket in Jira or GitHub, and the same bundle is what a sandboxed fix run reads before it opens a pull request with a repro test that fails before the fix and passes after.

Where they are genuinely good

Pastel is ahead of Snag in three places. A comment left on a video is tied automatically to the timestamp it was left at, and there is no Snag equivalent of that at all. Six project destinations are listed on the Pastel plans page, where Snag connects Jira and GitHub. And nineteen tools sit behind a Pastel MCP server that any coding agent can drive, where the Snag server only hands out evidence. If the job is getting a client to sign off on work, that is what Pastel is for, and none of this table argues against it.

Where Snag is different

A review closes when the person who asked for it is satisfied. A bug closes when there is proof it is gone. Producing that proof is the whole of what Snag does. The evidence that filled the ticket is the evidence the coding agent reads: it reproduces the failure in a disposable copy of your repo, writes the test that fails on the old code and passes on the new, and opens a pull request for a person to review. Nothing merges without a human, and the sandbox is destroyed when the run ends. Runs that end without a pull request are billed at half, so what you pay for is a work product rather than an attempt, and the first 5 fixes on any plan cost nothing.

Which one to pick

Choose Pastel when the thing under review is a design, a page or a video, and the people commenting are clients rather than engineers. Choose Snag for the other half of the job: the defect somebody hit in the running product, where whoever takes it on is going to open an editor, and where the useful answer is a reviewed change to the source rather than one more comment to reply to. Plenty of teams keep both, because the two jobs barely overlap.

Snag and Pastel, side by side

Every Pastel cell links to the page we read it on.

Questions people actually ask

Is Snag a drop-in replacement for Pastel?

It depends on what you are reviewing. Pastel is a review canvas: you share a link, and anyone can leave in-line comments on a live page or against a timestamp in a video, without an account. Snag asks a smaller question and answers it much further down the line: what broke, on which screen, with which console errors and which failed requests. That is captured in one click, filed in Jira or GitHub, and handed to a fix run that comes back with a pull request. The capture itself is masked in the browser before anything uploads — inputs are never recorded. If the work is getting a client to sign off on a design, neither product stands in for the other.

How does Snag's pricing compare to Pastel's?

The Pastel Pro plan costs $29 a month when billed annually, which their plans page states as $350 for the year, and $35 a month when billed monthly, with $24 for each additional user. Snag prices the workspace rather than the people in it: one figure covers capture, routing and the MCP server for everybody who touches it, and a fix plan adds a monthly allowance of fix-runs on top of that. Today's numbers are on the pricing section of the homepage.

Pastel runs an MCP server too. What does the Snag one do differently?

Reach, and who runs the coding agent. Pastel exposes nineteen tools and describes the loop they enable: a comment goes through to a fix in your code and a reply on the comment, in a single pass, driven by the coding agent you set up and pay for. The Snag server is deliberately narrow: it reads evidence out and writes nothing back, because the writing half is the fix run, and the fix run ships with the product instead of being something you assemble. Either route is fine here — point your own agent at the server, or leave it to the built-in run.

What does a Snag fix run hand back at the end?

Three things, and the middle one is the whole point: the diff, a repro test that fails on the code as it stands and passes on the code with the change, and the capture the agent worked from, linked so a reviewer can check the reasoning instead of taking it on trust. All of that happens in a disposable copy of your repo, destroyed when the run ends. Then it stops and waits for you — Nothing merges without a human. A run that never reaches a pull request is billed at half, and the first 5 fixes on any plan cost nothing.

Where these details come from

Every Pastel detail on this page links to the Pastel page we read it on, with the date we read it. Something out of date? Email admin@dvcllc.io and we will fix it.