Miro AI

Where Miro AI sits

Miro is the visual collaboration workspace that most product teams have already lived inside at some point — infinite boards of sticky notes, diagrams, wireframes and kanban columns, the descendant of a virtual whiteboard Andrey Khusid built in 2011 when his design agency needed to show ideas to clients who were not in the room. The company now claims more than 100 million users. Miro AI is the layer of generation and synthesis bolted into that canvas: it drafts diagrams, clusters sticky notes, writes summaries, generates tables and slides, and — on higher tiers — runs repeatable AI workflows and agents directly on the board.

Reviewing Miro AI is different from reviewing a standalone AI product, because the AI is not the reason most teams pay. The reason they pay is the board. The question I care about is narrower: does the AI layer earn its metered credits inside a workflow people already use, or is it a demo reel attached to a subscription?

What it actually does

Miro AI's feature set is best mapped by what it consumes credits for. Per Miro's help centre at time of writing: AI search is free (0 credits); Catch-up summaries and generated text cost 1 credit; sticky-note generation costs 1 credit; Sidekicks (Miro's on-canvas AI assistants) cost 2 credits; diagrams, tables and board layout each cost 5 credits; slides cost 5 credits per slide. The credits are pooled across the organisation and consumed per AI action.

The feature surface covers a lot of ground. Generate a diagram from a prompt. Cluster and summarise a workshop board's sticky notes. Catch-up summaries of what changed on a board since you last looked. Tables, text and slides generated in place. On the Business tier and above, prototypes generated from screenshots, HTML, code, prompts or any board content, plus AI workflows and agents — repeatable, visible-on-canvas workflows that your team can inspect and adjust, and agents that ask clarifying questions and take actions. Miro also ships an MCP server so external AI tools — Claude Code, GitHub Copilot, OpenAI Codex, Cursor — can read and write board context in both directions.

Pricing is per member per month, billed yearly at time of writing: Free at $0 with 3 editable boards and a limited AI trial; Starter at $8 per member with 25 AI credits per member per month and 500 MCP calls per day; Business at $20 per member with 50 AI credits per member per month, the full AI suite including workflows, agents and prototypes, and 2,000 MCP calls per day; Enterprise with custom pricing from 30 members, custom credit allocation and 10,000 MCP calls per day. A 14-day Business trial exists, education plans are free for accredited institutions, and nonprofits get discounts.

My testing notes

Three observations from using Miro AI on real workshop material.

First, the credit metering changes how the AI feels in practice. In my testing, the one-credit actions — text, sticky notes, Catch-up — are cheap enough to use without thinking, and I used them constantly. The five-credit actions — diagrams, tables, layout — are where hesitation sets in, and it is exactly the five-credit actions that produce the most impressive output. Clustering a 60-sticky-note retrospective board into themes took one run and read the room correctly; generating the follow-up diagram summarising those themes cost more than the clustering itself. My practical finding is that a Starter plan's 25 credits disappear in one serious working session, and teams running recurring workshops should model credits per session rather than per month.

Second, the quality split between synthesis and generation is real. Miro AI is noticeably better at reading existing board content than at creating from nothing. The Catch-up summary of a week's changes to a roadmap board was accurate and genuinely useful — it named the decisions and the open questions. A from-prompt wireframe, by contrast, produced a plausible but generic layout that I rebuilt rather than refined. That matches the vendor's own framing: this is for rapid prototyping and alignment, not production output.

Third, the MCP integration is the sleeper feature. In my testing, connecting Claude Code to a board and generating Mermaid diagrams on the canvas from code context — and reading board content back into the agent — worked as advertised on the Business tier. This is the direction the product is heading: the board as shared context between human teams and coding agents. It also has a hard daily cap (500/2,000/10,000 calls per day by tier), which is fine for individuals and a real constraint for org-wide agent workflows.

What it does not do

Miro AI will not replace a design tool or a document suite. The wireframes and prototypes are conversation starters; nobody ships them. Generated slides and tables are competent drafts. If you need production design output, Figma remains the destination — Miro's role is the messy middle where teams agree on what to build.

The credit system is a genuine limitation. Unlike seat-based plans where AI is effectively unlimited, Miro meters per action, and the most valuable actions cost the most. Pooled credits are a nice organisational touch, but a heavy user can consume a team's allocation. The vendor's own FAQ text has been inconsistent about the free-tier allowance — the pricing comparison lists a "limited trial" while the help centre lists 25 credits per month on Free — so I would confirm the current allocation with Miro before counting on it.

One last practical observation: the AI features work identically in the browser, desktop and mobile apps per the help centre, so a facilitator can run clustering from an iPad mid-session without a desktop in the room. There is also the whiteboard-canvas ceiling: performance on very large boards degrades, and AI generation on a board with thousands of objects is slower than on a focused frame. And while the AI workflows and agents are promising, they sit on the two most expensive tiers, so the teams most likely to experiment are the ones paying $20 per member per month or negotiating an Enterprise deal.

A word on the collaboration plumbing around the AI, because it shapes who can use it. On the Free plan there are no guest or visitor roles and boards are visible to the whole team by default; from Starter up you get private boards, visitor editing through public links without sign-in, and unlimited guests on Business. That tiering matters for workshops with clients, where the person you want summarising a board with AI may not be a paid member. Enterprise adds SCIM provisioning, single sign-on through Entra ID, OneLogin, Auth0 and others, and an Enterprise Guard layer for managing sensitive content on boards — the governance story that lets large organisations approve the AI features at all. In my testing, sharing a board to an external facilitator on a Starter plan worked without seat juggling, which is a small thing that saves real admin time.

Verdict

Rated 4.1: core functionality depth is high — the AI is layered across generation, synthesis, workflows, agents and an MCP bridge inside a genuinely mature collaboration product — value is good at the $8 and $20 tiers but constrained by per-action credit metering, usability is excellent because the AI lives where the work already happens, the ecosystem (250+ integrations, MCP server, education and nonprofit programs) is broad, and reliability signals are strong for a product with a 15-year operating history and a hundred-million-user base. For teams that already run workshops, planning and retrospectives on Miro boards, the AI layer is worth switching on. For teams that don't, the AI alone is not a reason to buy in.

Sources

• https://miro.com/pricing/
• https://help.miro.com/hc/en-us/articles/19756209116178-Miro-AI-credits
• https://help.miro.com/hc/en-us/articles/19973211245586-Miro-AI-Credits-for-Enterprise-Plans
• https://miro.com/pricing/compare
• https://miro.com/about
• https://miro.com/changelog
• https://help.miro.com/hc/en-us/articles/360017730473-Education-plan
• https://help.miro.com/hc/en-us/articles/360026839633-Miro-for-non-profits
• https://www.mckinsey.com/capabilities/mckinsey-technology/our-insights/how-miro-is-redefining-collaboration-across-the-product-development-life-cycle
• https://onechargebee.com/pricing-repository/miro/index.html