Eraser turns prompts, code, Git repositories and uploaded files into editable architecture diagrams, flowcharts, ERDs and design docs for engineering and architecture teams.
Used this tool? Rate it
Used this tool? Rate it
Eraser is an AI diagram generator and documentation workspace for software and architecture teams, operated by Eraser Labs, Inc. from a San Francisco address. The company calls Eraser an AI tool for technical design and documentation, and it also built the DiagramGPT generator on OpenAI models.
Eraser's tagline is "AI for diagrams that matter", and the company says it creates production-ready diagrams in seconds using AI. Prompts, code snippets, Git repositories and uploaded files go in; editable architecture diagrams, flowcharts, entity relationship diagrams, sequence diagrams and BPMN diagrams come out.
Every Eraser file consists of a single infinite canvas and a note editor, so a diagram and the design doc that explains it live in the same place. That pairing matters because AI output is not a flat picture: it lands on the canvas as editable diagram code that can be refined by prompt, by code or by hand.
The product has changed a lot since launch. In 2021 Eraser was a year-old startup centered on a collaborative digital canvas for whiteboarding and note-taking, and it raised $4 million in seed funding led by Caffeinated Capital. The current product layers Git-connected codebase diagrams, a CI bot, an API and an MCP server for AI agents on top of that canvas.
Because the AI writes diagram code rather than pixels, a generated result can be corrected line by line, regenerated with a new prompt, or adjusted visually, and all three paths edit the same diagram.
Eraserbot uses credits only when a diagram is actually created for a PR with relevant code changes or when someone manually regenerates it, so a PR whose changes do not affect the diagram does not consume credits.
Eraser is aimed at people who need an architecture diagram tool tied to real systems rather than a general drawing canvas.
Eraser is not for minors: the terms require users to be over 18 and do not permit anyone under 18 to use the service.
| Plan | Price per member (annual billing) | Monthly billing | Files | AI diagrams | Version history | Other inclusions |
|---|---|---|---|---|---|---|
| Free | $0 | $0 | 3 | 3 or 5 (pages differ) | 7 days | Unlimited guests |
| Starter | $15 | $20 | Unlimited | 40 | 90 days | Private files, 20 custom icons, API calls |
| Business | $45 | $60 | Unlimited | 250 | Unlimited | SAML SSO, 100 custom icons, reference diagrams |
| Enterprise | Custom | Custom | Unlimited | Unlimited | Unlimited | Flexible deployments, customer success manager, ACH and wire transfer, usage reporting |
The pricing page shows the Free plan at $0 per member per month with 3 files, 3 AI diagrams and 7-day version history. The plans documentation instead says Free teams can generate up to 5 AI diagrams, so the two official pages disagree on the Free AI allowance.
The privacy policy does not address AI model training, and the review-platform pages checked this round were blocked, so no independent rating sample is reported here.
Yes, there is a Free plan at $0 with 3 files and 7-day version history. Its AI allowance is listed as 3 AI diagrams on the pricing page and up to 5 in the plans documentation.
Eraser says neither it nor OpenAI will train models on your data, although Eraser may analyze usage to improve its AI feature. The privacy policy itself does not address model training.
Yes. Results are stored as diagram code, so you can change them with further prompts, by editing the code, or by hand on the canvas.
Yes. File-to-diagram converts Visio and draw.io files, screenshots, spreadsheets and slide decks into native, editable diagrams, up to 10 MB per file.
Both exist. The API is limited to paid plans and requires usage-based pricing, while the MCP server is free to use within each plan's AI credit limits.
