Comparison

Writhink vs BugHerd: a feedback layer, not a task system

BugHerd is a full website QA system — point-and-click pins, automatic browser and OS capture, an integrated Kanban board, a long roster of connectors and an MCP server of its own. Writhink is deliberately much smaller: it collects client feedback and turns it into one implementation prompt your AI assistant can run. If you need the board and the Jira link, BugHerd is the answer, and this page says so.

No credit card. Your client doesn’t need an account.

01

What BugHerd is good at

You pay BugHerd for internal members, not for the people leaving feedback, and none of your clients need an account. That is the strongest client-access position in this category, and for an agency with a long stakeholder list it is materially cheaper than any seat-counted alternative.

The capture is automatic and thorough: screenshot plus browser, OS, resolution and the selected element, so a report arrives actionable without a follow-up conversation. The Kanban board means BugHerd can be the task system rather than a feeder into one. It also does Figma prototypes, PDFs, images and documents, text-edit suggestions, and short video or audio feedback, with browser extensions beyond Chrome and public feedback from logged-out site visitors.

Their AI story is the deepest of any incumbent here. BugHerd MCP is in beta, free on every tier with no upgrade required, and exposes a broad set of read and write tools across a long list of AI clients, with in-app setup for Cursor. BugHerd AI adds auto-title generation on all tiers, and auto-tagging plus similar-task detection on their top tier. Anyone telling you the incumbents have no agent story is out of date, and that includes anyone selling you Writhink's MCP server as a unique feature. It is not one.

02

The two MCP servers, side by side

Both let an AI agent read feedback and write back. They are not the same shape.

WrithinkBugHerd
Tools exposed7, all on the feedback side: read a round with its pin context, reply to the client, restatus, create a project, install the widget11 read and write: task detail, task listing, project metadata, status, priority, severity, due date, assignee, comments, task and project creation, project settings
StatusLive for every account, not gated by planBeta, free for all users, no tier restriction
ClientsAny MCP-compatible client, including Claude Code, Claude Desktop and CursorClaude Code, Claude AI, Cursor, ChatGPT, VS Code with Copilot, Windsurf, Codex CLI, Gemini CLI, Zed, Onyx AI
AuthenticationOAuth 2.1; the agent acts as youNot covered in what we checked
What the agent readsClient comments with pin position, element anchor, page URL, viewport, device type and screenshotFull task detail with screenshot, URL and browser data
Known limitOne active organization at a timeAI features are still labelled beta

Writhink's tool-by-tool reference and its authorization model are on /visual-feedback-for-ai-agents. Competitor details checked on 2026-09-01. Pricing and features change — check their site before deciding.

03

The difference is not the tool count

More tools beat fewer on paper. What matters more is what is inside the payload.

  • BugHerd hands an agent tasks that are already structured: a title, a status, a priority, browser data, a screenshot. If your feedback comes from testers, that structure is most of the work done.
  • Writhink hands an agent client sentences, and then does something about the fact that they are client sentences. "This bit looks broken on mobile" becomes an item naming the page, the element and the breakpoint, with the screenshot attached.
  • The generated prompt has no equivalent on the other side: one text block covering every open comment, duplicates merged, grouped by page and element, prioritized, with requests that stay ambiguous flagged rather than guessed at.
  • BugHerd's board is where work lives. Writhink assumes work lives in your editor and your AI assistant, so it tracks four statuses and gets out of the way.
  • Nothing stops you running both: Writhink for the client review, BugHerd for internal QA. They are not competing for the same seat.

04

What each side charges for

Two different meters, and one of them is temporary.

WrithinkBugHerd
Free tierFree during the open beta, no cardNone; trial only
Entry planNo tiers during the betaStandard $50/mo, or $42 billed annually, for 5 members
Higher tiersNo tiers during the betaStudio $80/mo or $67 annually for 10 members; Premium $150/mo or $125 annually for 25
Extra seatsNot counted during the beta$8 per member per month, or $6.60 on annual billing
Clients and reviewersNo account, not countedUnlimited on every plan; only internal members are metered
StorageNot metered todayCapped, from 10GB on Standard
AI featuresPrompt generation and MCP on every accountMCP free on every tier; auto-tagging and similar-task detection on the top tier

Competitor details checked on 2026-09-01. Pricing and features change — check their site before deciding.

05

Which one should you actually run

Writhink fits when

  • Your reviewers are clients, not testers, and what they write is an opinion rather than a bug report.
  • An AI assistant does the implementation and you want the input written for it.
  • You do not want another board. Your work already lives in a repository and an editor.
  • You review websites and web apps — and PDFs and image sets, on a still-green side of the product — and would rather not watch a storage cap while the screenshots pile up.
  • You want to start today without a card and without picking a tier.

BugHerd fits when

  • The feedback has to reach Jira, Linear, GitHub, Asana, ClickUp, Trello, Slack or one of a long roster of other connectors. Writhink integrates with nothing — not fewer connectors, zero.
  • The tool needs to be the task system: a board, assignees, severity, due dates.
  • You need automatic browser, OS and resolution capture on every report, or video and audio feedback.
  • Your reviewers use Firefox, Safari or Edge. BugHerd has extensions across major browsers; Writhink's is Chrome only.
  • Your stakeholder list is long and unlimited clients on every plan is worth paying for. Writhink's beta terms will not stay free forever.
  • You want a mature product with years of production use. Writhink is new and cannot offer that.

06

What Writhink captures, exactly

So nobody buys it expecting diagnostics: a Writhink comment carries the pin position, the element anchor, the page URL, the page index, the viewport size, the device type — desktop, tablet or mobile — and an automatic screenshot of the pinned spot. Attachments are allowed, up to five files of 10 MB each.

It does not capture the operating system, the browser or the handset model. No console logs, no network requests, no session replay of what your client did. Device type is a category, not a fingerprint: you will learn that a complaint came from a phone at a 390-pixel viewport, never that it came from a particular model or browser version. If a report needs any of that to be reproducible, Writhink is the wrong tool and BugHerd or a dedicated QA product is the right one.

Before you count on it

Writhink is in open beta: everything described here is available and free, no credit card. Paid plans are announced but not purchasable yet. The one limit that applies today is five AI prompt generations a month per organization.

FAQ

Questions people actually ask

Does Writhink have a Kanban board?
No. There are four statuses — open, in progress, in review, resolved — and a list. If you want swimlanes, assignees, severity and due dates, that is BugHerd's territory and it is a real reason to stay there.
Can Writhink push feedback into Jira or Linear?
No. Writhink has no native integrations with any tracker, builder or chat tool. Feedback leaves two ways: as the generated implementation prompt, copied or downloaded as a .md, or through the MCP server so an agent can read it and act. If a connector is a hard requirement, BugHerd lists a long roster of them and Writhink lists none.
Which MCP server is better?
For task management, BugHerd's: its tools cover status, priority, severity, due dates and assignees, and the table above has the counts and the date we checked them. For client feedback aimed at an AI that will write the code, Writhink's payload is the one built for that job — comments rewritten into clear items, grouped and prioritized, with unclear requests flagged rather than invented. Neither of us can claim MCP as a differentiator any more; the argument is about what rides on it.
Do clients need an account with either?
No, with both. BugHerd's clients are unlimited on every tier and submit through shared project links without logging in. Writhink asks a guest for a display name; a browser-stored token proves ownership so they can edit and delete their own comments and reply publicly.
Is Writhink a bug tracker?
No, and it is not trying to become one. It is a layer between a client and an AI coding assistant. The moment your process needs assignment, severity and a board, you have outgrown it.

Start

Send one link. Get feedback you can build from.

Your clients comment right on the page — no account, no screenshots, no voice notes. Writhink sorts the comments and writes the prompt for your AI.