Mermaid is the commercial platform built by the team behind the open-source Mermaid.js library, turning prompts, documents and Markdown-style syntax into version-controllable diagrams you can co-edit, embed in your IDE, or generate through an MCP server.
Used this tool? Rate it
Used this tool? Rate it
Mermaid — the product at mermaid.ai, operated by Mermaid Chart Inc. — is a commercial diagramming platform built directly on top of the open-source Mermaid.js library. Its core proposition is that diagrams should be written, generated or prompted rather than drawn: you describe a system in text, in a Markdown-like syntax or in plain language, and the diagram lays itself out.
Getting the identity right matters here more than for most tools, because three different things share this name. They are related but not interchangeable:
github.com/mermaid-js/mermaid. It renders diagrams from text and is what GitHub, GitLab and countless documentation tools embed natively. It is free, and always will be.The relationship is verifiable rather than merely claimed. The GitHub repository mermaid-js/mermaid sets its homepage field to https://mermaid.ai/open-source/, so the open-source project itself points at this domain — bidirectional confirmation from the code hosting side. Separately, the former product domain mermaidchart.com now redirects to mermaid.ai, confirming these are one entity under a newer domain rather than two competing products. The site states its own position plainly: "From the team behind Mermaid.js," with a note that every diagram made on the platform "keeps Mermaid's open-source roots growing."
The origin story is unusually well documented for a tool this size. TechCrunch reported in March 2024 that founder Knut Sveidqvist created Mermaid after losing a Microsoft Visio file and having to redraw everything — on the way home he reasoned there had to be a better way, and that Markdown was starting to take hold, so "can't we have this as text?" He started the project that evening while his children watched The Little Mermaid, which is where the name comes from.
The platform's defining design choice is that AI generation, code authoring and visual editing are three views of one artifact rather than three separate products.
Create with AI. Describe what you want, upload a document, or paste a transcript, and get a first-draft diagram you can shape. The platform states that Mermaid AI holds context across the whole conversation, so the diagram grows as your thinking does rather than resetting each turn. It also claims to suggest the diagram shape that fits what you are building — flowchart, timeline, requirement diagram, sequence — instead of just producing whatever you asked for literally.
Build with code. Write in Mermaid syntax and watch the layout resolve itself. The bidirectional navigation is the genuinely useful part here: click any node to jump to its code, or click the code to find its shape on the canvas, so you always know where a change lives. For anyone maintaining a large diagram, that mapping is the difference between editable and disposable.
Edit visually. Prompt inside the editor, adjust with a click, snap layouts with node alignment, and reattach edges without rewriting a line. TechCrunch noted this visual editor was added specifically because a text-based tool carries a relatively high barrier to entry — and that it is opt-in, so users who prefer text can stay in text and switch between interfaces at will.
The site surfaces a broad type list rather than a generic canvas: flowchart, sequence diagram, Gantt chart, ER diagram, timeline, mind map, org chart, block diagram, state diagram, Git graph, user journey, class diagram, pie chart and quadrant chart. Architecture and C4 diagrams also appear in the role-specific recommendations.
The type list is the real product surface. A general drawing tool gives you shapes; Mermaid gives you diagram grammars, each with its own layout rules — which is why a sequence diagram automatically arranges lifelines rather than requiring you to place them.
This is where the platform separates itself from browser-only diagramming tools. Diagrams render in your IDE, embed in docs and wikis, and — notably — can be generated straight from Claude, VS Code and Cursor through the platform's MCP server, which can generate, validate and render diagrams with no manual syntax needed. The site also lists GitHub Copilot and Confluence among supported environments, and a dedicated VS Code integration for turning infrastructure and systems into diagrams from inside a codebase.
For anyone already working with AI coding assistants, MCP support is the most consequential feature on this list: it means diagram generation becomes a tool call in an existing agent workflow rather than a context switch to another tab.
Co-edit live, share with a single link that requires no login for the viewer, and walk people through diagrams without exporting to a second tool. The no-login share link matters practically — it removes the most common friction point when sending a diagram to someone outside the team.
The platform advertises secure access controls, trusted encryption, SOC 2 compliance, and SSO that fits existing setups. SSO is gated to the Premium tier and above per the published plan comparison. Note that SOC 2 compliance is stated on the marketing page; the site does not link a report or attestation letter at that location, so treat it as a vendor statement rather than a verified artifact until you request the report through sales.
Documentation that stays current. This is the strongest case and the one the tool is architecturally suited to. Because diagrams are text, they live in version control alongside the code they describe, diff meaningfully in a pull request, and can be updated by whoever touches the system rather than only by whoever owns the drawing file. A user testimonial on the site captures the motivation precisely: the tool "solved a major portability problem for us, making charts behave like code."
Architecture and system design. The site's role mapping points system architects at C4 and ER diagrams, tech leads at architecture, sequence and C4, and DevOps engineers at Git graphs and flowcharts for CI/CD pipelines and infrastructure.
Explaining a system to someone who did not build it. Onboarding, handover and stakeholder alignment. One testimonial reports that what used to take 30+ minutes in GUI tools takes under five.
AI-assisted codebase visualization. With the MCP server or the VS Code integration, you can point an assistant at a codebase and get a diagram of its actual structure. One quoted user describes using AI to analyze their software in order to visualize complex logic instantly — a workflow that did not meaningfully exist before agent tooling.
Project and process planning. Gantt charts, timelines and user journeys serve project managers, business analysts and consultants — the audiences the company has explicitly said it wants to reach beyond its developer base.
Teaching and thinking out loud. Mind maps, timelines and flowcharts for educators and what the site calls "visual thinkers." The homepage prompts include "Organize my thoughts," "Explain a complex idea" and "Teach a lesson."
Step 1 — Decide whether you need the platform at all. This is genuinely the first question, and the vendor answers it honestly on its own site: Mermaid syntax does not require a paid plan. If you only need to render a diagram inside a GitHub README or a documentation site, the open-source library already does that at no cost. The platform earns its keep through AI generation, collaboration, storage, presentations and administration — not through access to the syntax.
Step 2 — Start free and understand the ceiling. The Basic tier is free and includes 3 diagrams (the comparison table lists up to 6), limited diagram size (capped at 60), 15 AI credits, and one team and project creation. That is enough to evaluate the workflow, not enough to run on.
Step 3 — Pick your entry point. Type a prompt, upload a document, paste a transcript, or start from a diagram type. If you already think in Mermaid syntax, go straight to the code panel; if you do not, start from AI or from one of the homepage prompts such as "Design a system" or "Map out my team."
Step 4 — Refine in whichever view suits the change. Structural changes are usually faster in code; cosmetic and layout changes are usually faster on the canvas. Because clicking a node jumps to its code and vice versa, moving between the two costs almost nothing.
Step 5 — Wire it into where the work actually happens. Connect the MCP server if you use Claude, VS Code or Cursor; install the VS Code integration if you want diagrams generated from your codebase; use the GitHub Copilot or Confluence integrations if your documentation lives there. A diagram tool that requires you to leave your editor tends to be abandoned.
Step 6 — Share by link, present in place. Share with a link that needs no login for the viewer, and present directly rather than exporting to slides.
Budget AI credits annually, not monthly. The AI allowances are published per year, not per month: 15 on Basic, 300 on Plus, 2,000 on Premium, and "Unlimited*" on Enterprise (the asterisk is the vendor's own, indicating conditions apply). A 300-credit annual allowance is roughly 25 AI generations a month — comfortable for occasional use, thin if AI is your primary authoring method. If you intend to prompt your way to most diagrams, Premium is the realistic starting tier.
Keep diagrams in source control. The main structural advantage over GUI diagramming tools is that a Mermaid diagram is text, so it belongs in the repository next to the thing it documents. A diagram that lives only in a SaaS account will drift out of date exactly as fast as a Visio file did.
Use the type list as a design prompt. If you are unsure whether something is a sequence diagram or a state diagram, that uncertainty usually means you have not decided what question the diagram answers. The platform's suggestion feature helps, but choosing deliberately produces better diagrams than accepting the first shape offered.
Check the diagram size cap before you start on the free tier. Basic limits diagram size to 60 — a real ceiling for architecture work. Hitting it midway through a complex diagram is a frustrating way to discover the limit.
Expect rough edges on complex layouts, and know where to check. The open-source renderer that powers the platform has a substantial backlog of layout and styling issues, some long-standing. Before committing to a specific visual result, it is worth searching the project's GitHub issues for your diagram type — this is the honest way to set expectations, and the information is public.
Annual billing is the headline price. Both Plus and Premium prices are quoted per user per month billed annually. The FAQ indicates monthly billing is addressed separately, so confirm the monthly rate before assuming the advertised figure applies to a month-to-month commitment.
Viewer seats change the team math. Premium includes unlimited free viewer seats. For an organization where a handful of people author diagrams and many more only read them, that provision can matter more to total cost than the per-seat price itself.
Engineering teams that already treat documentation as code are the natural fit — the version-control story, the Git graph support and the IDE integrations all serve this group first.
Developers using AI coding assistants get disproportionate value from the MCP server, since it folds diagram generation into an agent workflow they already run.
Technical writers and documentation specialists benefit from diagrams that survive edits and stay in the same repository as the prose they illustrate.
Non-technical roles the company is deliberately courting — project managers, business analysts, consultants, HR managers, educators — are served by the visual editor and AI generation, which exist precisely because the text-first approach was a barrier. Whether that lands depends on the individual; the underlying model is still a diagram grammar.
Who should look elsewhere: anyone whose primary need is free-form visual design or brand-styled illustration (a canvas tool like Figma or Miro fits better); anyone who needs extensive template libraries or pixel-level control over appearance; anyone who needs only to render a diagram in a README, in which case the free open-source library is sufficient; and anyone unwilling to accept the layout constraints of automatic diagram engines.
Mermaid is a web platform, with the workspace at mermaid.ai and sign-in via Google or GitHub as well as email. Its reach beyond the browser is through integrations rather than native applications:
The open-source library extends this reach considerably further, since Mermaid.js is supported natively in GitHub and GitLab among others — a point Open Core Ventures founder Sid Sijbrandij made when the seed round was announced.
An API is referenced in the site footer, but the link routes to a Google Forms sign-up rather than to public documentation, which suggests it is gated or in a request-access phase rather than generally available.
Four tiers, all quoted as billed annually. These figures were read directly from the rendered pricing page.
Basic — free. For getting started and simple flows. 3 diagrams (the comparison table states up to 6), limited diagram size (60), limited AI at 15 credits, and 1 team and project creation. Co-editing and external sharing are included. No viewer seats, no SSO.
Plus — $10 per user per month, billed annually. For individuals who diagram often. All Basic features plus unlimited diagrams, limitless diagram size, and 300 AI credits per year. Still no viewer seats and no SSO.
Premium — $20 per user per month, billed annually. Marked "Most Popular." For advanced users and teams. All Plus features plus 2,000 AI credits per year, unlimited free viewer seats, unlimited teams and project creation, and SSO.
Enterprise — custom, billed annually. For organizations that need scale and control. All Premium features plus custom contract and invoicing, secure diagram ownership management (IP transfer to admin), a dedicated Customer Success Manager, user consolidation, and AI credits listed as "Unlimited*" with the vendor's own asterisk.
What the free tier does not restrict. Worth stating plainly because it is unusual: the vendor's FAQ addresses head-on whether a paid plan is required to use Mermaid syntax. The open-source library remains MIT licensed and free regardless of plan. You are paying for the hosted platform, not for the diagramming language.
Billing questions the FAQ covers but does not answer on the public page. Free trials of Plus and Premium, mid-cycle plan switching, monthly versus annual billing, accepted payment methods, and whether AI usage costs extra are all listed as FAQ topics on the pricing page, but the answers render behind accordion controls that did not expand to static text during capture. Confirm these directly with the vendor before committing — particularly the monthly billing rate, since every published price is annual.
Mermaid.js itself is the most important alternative and the one the vendor cannot avoid competing with. MIT licensed, 89,878 stars and 9,193 forks on GitHub as of capture, natively rendered by GitHub and GitLab. If you need diagrams in a README and nothing more, this is the correct answer and it costs nothing.
PlantUML is the closest philosophical competitor — also text-to-diagram, longer established, with broader UML coverage but a heavier toolchain and a less modern rendering aesthetic. The vendor maintains its own comparison page for it.
draw.io / diagrams.net is free, powerful and canvas-first. It wins on visual control and template breadth; it loses on version control, since the artifact is a file rather than readable text. Also covered by a vendor comparison page.
Lucidchart and Miro are the enterprise canvas incumbents. TechCrunch reported the company's stated ambition is not primarily to disrupt these — the longer-term target named was ServiceNow and similar workflow automation services, though the article noted that is "still quite a ways off."
Whimsical competes on speed and polish for product and design teams, with a third vendor comparison page devoted to it.
Excalidraw is the alternative for anyone who wants the hand-drawn, low-formality look — a different design intent entirely.
Layout control is the structural trade-off, and the evidence is public. Automatic layout is the reason text-based diagramming is fast, and it is also its main constraint: you get the engine's arrangement, not yours. The open-source project that renders these diagrams carries 703 open issues labelled as bugs or errors at time of capture, and the most-reacted ones are specifically about layout and styling control. Among the top items: flowchart arrow heads not being colored per linkStyle settings (open since February 2020, 76 reactions), subgraph label spacing missing left/right and bottom padding (open since January 2020, 62 reactions), multiline titles in flowchart subgraphs being overlapped by nodes (open since November 2022), and massive whitespace above and below graphs (open since April 2021). These are not obscure edge cases — they are the highest-voted issues in the project, several have been open for more than five years, and they describe exactly the cosmetic control that a design-oriented user expects.
The text-based approach has an acknowledged barrier to entry. This is not an outside criticism; TechCrunch reported that the team is "quite aware" that as a text-based tool there will always be a relatively high barrier to entry, and that the visual editor was added in response. If your team is largely non-technical, budget for the learning curve or plan around the visual editor from the start.
Independent review evidence is very thin. This is the honest state of the third-party record. Trustpilot's profile for mermaidchart.com shows 3.5 out of 5 from just 3 reviews, on an unclaimed profile — Trustpilot itself notes the company has no history of asking for reviews, so the sample may not be representative. Three reviews is far below the threshold at which a rating supports any quality conclusion, and it is reported here only as a matter of record, not as evidence of quality in either direction. G2 and Capterra list the product but both blocked automated retrieval, so no ratings from those platforms are quoted here rather than repeating figures that could not be verified firsthand. A Toolify directory entry shows 0 reviews.
The practical consequence: unlike a consumer product with hundreds of reviews, there is no meaningful aggregate user verdict available. The strongest available evidence about this product is its open-source footprint and its press coverage, not review scores.
AI credits are finite and the accounting is annual. Basic's 15 credits are an evaluation allowance. Plus's 300 per year averages around 25 a month. Only Premium's 2,000 supports AI as a primary workflow, and Enterprise's "Unlimited*" carries a vendor asterisk whose conditions are not stated on the pricing page.
SOC 2 is asserted, not documented on-site. The homepage lists SOC 2 compliance alongside encryption and access controls, but no report, attestation letter or trust-center link accompanies that claim at that location. For a procurement process, request the report directly.
Company scale and stage. This is a seed-stage company. TechCrunch documented a $7.5 million seed round announced March 2024, from Open Core Ventures, Sequoia and Microsoft's M12 fund along with about a dozen angel investors — a substantial round with credible backers, but it is early-stage funding, not the profile of an established enterprise vendor. The article also disclosed a correction: the team raised more funding and added investors between the reporter's first conversation and publication, and the story was updated accordingly.
Usage claims are vendor-reported. "Trusted by over 5M people and over 200k companies" appears on both the homepage and the pricing page as the company's own figure, with no independent audit. A related but distinct data point from TechCrunch is that CEO Andrew Firestone said the cloud version of the open-source project had 4 million users in the year before the March 2024 article. Treat the current 5M figure as self-reported.
API availability is unclear. The footer's API link leads to a Google Form rather than documentation, indicating a request-access arrangement rather than a generally available public API.
Age restriction. The privacy policy of Mermaid Chart Inc., effective September 13, 2024, states the service is not intended for use by anyone under 16 years of age.
Mermaid.js is the open-source library — MIT licensed, hosted at github.com/mermaid-js/mermaid, and rendered natively by GitHub, GitLab and many documentation tools. Mermaid, the product at mermaid.ai, is the commercial platform built on top of that library by the same team, adding an editor, AI generation, real-time collaboration, presentations, integrations and team administration. A third thing, the Mermaid Live Editor, is a free browser playground for testing syntax. The relationship is verifiable: the open-source repository's own homepage field points to mermaid.ai/open-source/.
No. The vendor's FAQ addresses exactly this question, and the open-source library remains MIT licensed and free. If your requirement is rendering a diagram in a README, a wiki or a docs site, the library alone covers it. The paid plans buy the hosted platform — AI generation, collaboration, storage, presentations, SSO and administration — not access to the diagramming language.
All prices are quoted per user per month billed annually. Basic is free: 3 diagrams (up to 6 per the comparison table), limited diagram size (60), 15 AI credits, 1 team and project creation, with co-editing and external sharing. Plus is $10: unlimited diagrams, limitless diagram size, 300 AI credits per year. Premium is $20 and marked most popular: 2,000 AI credits per year, unlimited free viewer seats, unlimited teams and projects, and SSO. Enterprise is custom-priced: custom contract and invoicing, secure diagram ownership with IP transfer to admin, a dedicated Customer Success Manager, user consolidation, and AI credits listed as "Unlimited*".
They are allocated annually rather than monthly — 15 on Basic, 300 on Plus, 2,000 on Premium. In practice, Plus's 300 per year is roughly 25 generations a month, which suits occasional AI use but not an AI-first authoring habit. If prompting is how you intend to build most diagrams, Premium is the realistic entry point. Whether AI usage incurs extra cost beyond the allowance is listed as an FAQ topic on the vendor's pricing page but the answer did not render as static text during capture, so confirm it directly.
Yes. The platform provides an MCP server that lets you generate, validate and render diagrams directly from Claude, VS Code and Cursor with no manual syntax needed. There is also a dedicated VS Code integration aimed at turning infrastructure and systems into diagrams from inside a codebase, plus listed support for GitHub Copilot and Confluence. For teams already working with AI coding assistants, this is the feature that makes diagramming part of an existing workflow rather than a separate task.
Not necessarily, though the honest answer is that this tool was born text-first. You can describe a diagram in plain language, upload a document or paste a transcript and get a first draft. A visual editor lets you adjust by clicking, align nodes and reattach edges without writing syntax. That said, TechCrunch reported the team acknowledges that a text-based tool carries a relatively high barrier to entry — the visual editor exists precisely because of it, and it is opt-in rather than the default posture.
Less than in a canvas tool, and this is the main trade-off. Automatic layout is why the tool is fast; it also means you accept the engine's arrangement. The evidence is public and specific: the open-source renderer had 703 open bug-labelled issues at time of capture, and the highest-voted ones concern exactly this — arrow head coloring not following linkStyle (open since 2020, 76 reactions), subgraph label spacing, multiline subgraph titles overlapped by nodes, and excess whitespace around graphs. If precise visual control is a requirement, check the project's issue tracker for your diagram type before committing.
There is not enough independent review data to answer this properly, and it would be misleading to pretend otherwise. Trustpilot shows 3.5 out of 5 from only 3 reviews on an unclaimed profile, with Trustpilot itself noting the company has never solicited reviews — three reviews cannot support a quality conclusion. G2 and Capterra list the product but blocked automated retrieval, so their figures are not quoted here. The more meaningful signals are structural: the underlying open-source project has 89,878 GitHub stars and 9,193 forks, and the commercial entity raised a $7.5M seed round covered by TechCrunch.
The operating entity is Mermaid Chart Inc. Founder Knut Sveidqvist created the open-source project after losing a Visio file and having to redraw everything; he named it after The Little Mermaid, which his children were watching the night he started. In 2022 he was approached by Open Core Ventures — the venture firm of GitLab founder Sid Sijbrandij — which paired him with Andrew Firestone as CEO. TechCrunch reported a $7.5 million seed round in March 2024 from OCV, Sequoia and Microsoft's M12, plus about a dozen angels. So: credible backing and a genuinely large open-source footprint, but a seed-stage company rather than an established enterprise vendor.
The platform advertises secure access controls, encryption, SOC 2 compliance, and SSO — with SSO gated to Premium and above. Note that SOC 2 is stated on the marketing page without an accompanying report or attestation link, so request documentation through sales if it matters for procurement. Enterprise adds secure diagram ownership management with IP transfer to an administrator. The privacy policy of Mermaid Chart Inc., effective September 13, 2024, states the service is not intended for anyone under 16 years of age and provides a route to request data deletion.