ChartDB

Open-source database visualization, honestly reviewed. No marketing fluff, just what you get when you self-host it.

TL;DR

• What it is: Open-source (AGPL-3.0) database diagramming tool that imports your entire schema with a single SQL query — no database credentials handed over, no install required for the cloud version [1][5].
• Who it’s for: Developers, DBAs, and founding engineers who need to document or migrate a schema without setting up a full ER modeling suite. Also teams that want visual schema onboarding without paying Lucidchart prices [3][4].
• Cost savings: Lucidchart starts around $9/user/month and Visio is bundled with Microsoft 365 at ~$6/user/month. ChartDB self-hosted runs on a free Docker container with no table limits and no per-user charge [2][3].
• Key strength: The “Smart Query” import — one SQL command retrieves your entire schema as JSON, and the diagram renders in under 15 seconds. No connection strings, no credentials exposure, no slow crawling of information_schema by hand [1][5].
• Key weakness: The self-hosted edition is missing real-time collaboration, embedding, and automated live-sync — all of which are cloud-only. AI features require your own OpenAI API key on the self-hosted path, and the AGPL-3.0 license creates friction for anyone trying to embed this in a commercial product [2][3].

---

What is ChartDB

ChartDB is a web-based database diagram editor. The pitch is simple: run one SQL query against your existing database, paste the JSON output into ChartDB, and immediately get a visual ER diagram of every table and relationship. No connection strings are stored, no database credentials change hands, no agent sits in your VPC [1][5].

The project launched in 2024 and is now at 21,532 GitHub stars under AGPL-3.0. The codebase is TypeScript (99.8% of the repo) [5]. The team ships both a managed cloud at app.chartdb.io and an open-source version you can run locally via npm or Docker [README][2].

What separates it from most ERD tools isn’t the diagram editor itself — it’s the import flow. Most tools ask you to either (a) define tables manually by clicking through a UI, or (b) hand the tool a direct database connection. ChartDB’s “Smart Query” is a middle path: they’ve written database-specific introspection queries for each supported engine. You run it yourself, copy the result, and paste it in. The tool never touches your database directly [1][2].

Supported databases as of this review: PostgreSQL (including Supabase and Timescale variants), MySQL, SQL Server, MariaDB, SQLite (including Cloudflare D1), CockroachDB, ClickHouse, and Oracle [README][3].

---

Why People Choose It

The third-party coverage on ChartDB is thin compared to mature tools like Lucidchart — two verified SoftwareAdvice reviews [3] and one GetApp listing [4], plus a LogRocket tutorial [1] and a Medium overview [5]. That said, the user comments that exist are consistent:

The zero-credential pitch lands. The G2 quotes on the ChartDB homepage include one from a Tech Lead calling it “clean, powerful, and built for real engineering workflows,” and a Founder/CTO saying “great tool for visualizing and migrating databases” [3]. The LogRocket review calls out “instant schema visualization using a single query” as the standout feature, and the Medium overview notes that you don’t need to “share database passwords” even with the cloud version [1][5].

Onboarding new engineers. One SoftwareAdvice reviewer (Ohad, IT & Services, 51–200 employees) writes: “It instantly visualizes our Postgres schema, making it much easier for new engineers to understand the structure and relationships in our database. This reduces the time spent explaining schemas in meetings.” [3] The second reviewer (Lior, hospitality startup) says it helps him “design and implement database architecture visually, making it easier to collaborate and adapt as the project evolves.” [3] Both rate it 5/5 across features, ease of use, and value for money.

Versus Lucidchart and Visio. These are the tools ChartDB shows up next to on comparison pages [3][4]. Lucidchart is a general diagramming tool — it supports ER diagrams but doesn’t have database-native import. You’re drawing boxes manually or using a CSV import. Visio is similar. ChartDB’s Smart Query collapses what would take 30–60 minutes of manual diagram construction into a 15-second paste operation [1][2]. The trade-off is depth: Lucidchart has 2,248+ reviews, 54 integrations, and mature collaboration features. ChartDB is narrowly focused on database schemas and behind on collaboration tooling outside its cloud plan.

Versus dbdiagram.io. Not in the source data, but worth noting from context: dbdiagram.io uses a DSL (DBML) for code-first diagrams. ChartDB’s visual-first approach is meaningfully different — you don’t learn a DSL, you run a query.

---

Features

Based on the README, official docs, and third-party descriptions:

Core import and visualization:

• Smart Query: single SQL statement per supported database generates a full schema JSON [1][5]
• JSON paste import — no direct DB connection required [1][2]
• Renders tables, columns, data types, and foreign key relationships visually [1][3]
• Undo, redo, copy, paste, duplicate for diagram operations [README]
• Subject areas (grouping of tables) and notes [README]
• Version history — view schema diffs over time [3][4]

Export:

• SQL DDL scripts in the dialect of your choice (useful for cross-database migrations) [1][README]
• Export as image for documentation [1][4]
• DBML format [3]
• AI-powered DDL generation for migration scripts — paste your current schema, target a different engine, get the DDL [1][README]

AI features:

• AI-powered layout assistant: detects missing relationships, optimizes layout, groups related tables [3]
• AI diagram generator: describe tables in natural language, get an ER diagram [website]
• On cloud: AI included in Teams plan; AI Agent feature for generating diagrams from prompts [website]
• On self-hosted: requires your own OpenAI API key (or a custom OpenAI-compatible endpoint for local models like vLLM/Qwen) [README]

Collaboration (cloud-only features):

• Real-time multi-user diagram editing [3][4]
• Private diagram sharing, public diagram links [2][3]
• Embeddable diagrams for documentation sites [2]
• Automated sync: programmatically push database schema changes to diagrams via API without credentials [2]

Self-hosted specifics:

• No table limits (cloud free tier caps at 10 tables, Pro at 100, Teams at 200) [2]
• No built-in user account management [2]
• No team collaboration features — diagrams are local to your instance [2]
• No embedding or automated sync [2]
• Full source code access under AGPL-3.0 [README]

---

Pricing: SaaS vs Self-Hosted Math

ChartDB Cloud:

• Free: 10 tables, basic features, community Discord support [2]
• Pro: up to 100 tables — exact price not publicly listed on their current pricing page
• Teams: up to 200 tables, AI Assistant, real-time collaboration, private diagrams — SoftwareAdvice lists the paid starting price at $59/month [3]

The table cap on the free tier is the main friction point. A real production database with 10+ tables hits the limit fast.

Self-hosted:

• Software: $0 (AGPL-3.0) [README][5]
• VPS to run it: $5–10/month on Hetzner or Contabo
• OpenAI API key if you want AI features: pay-per-use, roughly $0.01–0.05 per diagram generation at current GPT-4o rates
• Your setup time: ~30 minutes for a technical user [1]

Lucidchart for comparison:

• Free: 3 editable documents, 60 objects per document
• Individual: $9/user/month
• Team: $10/user/month
• Enterprise: custom

Visio:

• Bundled with Microsoft 365 Business plans, or standalone ~$5/user/month
• No database-native import

Concrete math for a team of 5 engineers who need schema documentation:

On Lucidchart Team: 5 users × $10 = $50/month for a general diagramming tool where you manually draw database diagrams. On ChartDB Teams: $59/month flat for the database-specific tool with Smart Query import and AI layout. On ChartDB self-hosted: $6/month VPS with no user cap, no table cap, and full source access — minus the collaboration features.

For solo founders or small teams where one person owns schema documentation and exports are shared via image or DDL file, the self-hosted path is a clear win. For teams that need live collaboration and embedded diagrams, the Teams plan is competitive with Lucidchart on price and significantly better on database-specific UX.

---

Deployment Reality Check

LogRocket walks through the full self-hosted install [1], and it’s not complicated:

git clone https://github.com/chartdb/chartdb.git
cd chartDB
npm install
npm run dev

That’s it for a local dev server. For production Docker:

docker run -e OPENAI_API_KEY=<YOUR_KEY> -p 8080:80 ghcr.io/chartdb/chartdb:latest

Without the OpenAI key, the tool works fully for import, editing, and export — you just lose the AI-powered DDL generation and diagram layout features [README][1].

Custom inference server: The README supports pointing ChartDB at a custom OpenAI-compatible endpoint, so you can run it against a local vLLM server (example: Qwen/Qwen2.5-32B-Instruct-AWQ) and keep everything off OpenAI’s infrastructure [README]. The Medium overview notes this privacy angle — no credentials leave your server [5].

What can go wrong:

• The self-hosted version has no user authentication out of the box. If you expose it to the internet without putting it behind a reverse proxy with auth (Nginx + basic auth, Authentik, Cloudflare Access), anyone with the URL can use it [2].
• No built-in database — ChartDB is a stateless web app. Diagrams saved in the browser (local storage). There’s no persistent server-side storage in the community edition [2].
• Team collaboration requires the cloud. If your team needs simultaneous editing, you’re either on ChartDB Cloud or you’re using export/import to share diagrams manually [2].
• AGPL-3.0 is stricter than MIT. If you want to embed or distribute ChartDB as part of a commercial product, the AGPL requires you to open-source your modifications and the surrounding work in certain interpretations. Legal review required before commercial embedding [README].

Realistic setup time: 15–30 minutes for a developer who has Docker. The LogRocket tutorial covers it end-to-end and notes no major gotchas [1].

---

Pros and Cons

Pros

• Smart Query import is genuinely fast. 15 seconds from running the query to a rendered ER diagram, with no credentials exposure [1][5]. For schema documentation, this beats manually drawing diagrams by an order of magnitude.
• No table limits on self-hosted. The cloud free tier caps at 10 tables — production databases routinely have 50–200+. Self-hosting removes that cap entirely [2].
• Supports 8+ database engines. PostgreSQL, MySQL, SQL Server, MariaDB, SQLite, CockroachDB, ClickHouse, Oracle — most teams won’t hit a gap [README][3].
• Privacy-first architecture. No database password required. Smart Query output is just JSON — you paste it in, the tool never sees your credentials [1][2][5].
• AI-powered cross-database DDL export. If you’re migrating from MySQL to PostgreSQL, ChartDB generates the target DDL for you. For founders who inherited a SQL Server schema and want to move to Postgres, this is the feature [1][README].
• Local LLM support. Custom OpenAI-compatible endpoint means you can run AI features against a local model instead of sending data to OpenAI [README].
• Clean, fast UI. The Medium overview [5] and GetApp ratings [4] both point to ease of use as the standout quality. 5.0/5 on both platforms.

Cons

• AGPL-3.0 is not MIT. Commercial embedding requires a legal conversation. For founders who want to bundle a diagram tool into their SaaS, this is a real constraint [README].
• Real-time collaboration is cloud-only. The self-hosted version has no multi-user editing, no persistent diagram storage, no sharing links, no embedding [2]. For a tool pitched at “team collaboration,” this is a significant limitation in the free tier.
• AI features require OpenAI API key or custom endpoint. You don’t get AI out of the box on self-hosted — you set it up yourself [README][1].
• No built-in auth or persistence. Self-hosted ChartDB stores diagrams in browser local storage. No server-side database, no user accounts, no diagram history across browsers or devices [2].
• Public beta. The README labels the project as “currently in Public Beta” [README]. For production documentation workflows, that’s worth flagging.
• Thin review base. Two verified reviews on SoftwareAdvice [3], two on GetApp [4]. Not enough signal to assess long-term reliability, support quality, or edge cases at scale.
• Pricing opacity. The exact Pro plan price isn’t surfaced in the scraped data — SoftwareAdvice lists Teams starting at $59/month [3], but the full pricing structure isn’t clearly documented in the sources available.

---

Who Should Use This / Who Shouldn’t

Use ChartDB if:

• You need to document or understand a legacy database schema quickly — the Smart Query saves an afternoon of manual diagram work.
• You’re migrating between database engines and need generated DDL for the target dialect.
• You’re a solo developer or small team (2–3 people) who doesn’t need live collaboration and wants self-hosted with no table limits.
• Data sovereignty matters and you can’t share credentials with a third-party tool.
• You want AI layout assistance but want to run it against your own model endpoint.

Skip it (use the cloud version or wait) if:

• Your team needs real-time collaborative editing — self-hosted won’t give you that [2].
• You need persistent, server-side diagram storage across multiple users and sessions [2].
• You’re embedding this in a commercial product — review AGPL-3.0 implications first [README].

Skip it entirely (use Lucidchart) if:

• You need a general-purpose diagramming tool for more than database schemas — flowcharts, org charts, wireframes, infrastructure diagrams.
• Your org already pays for Microsoft 365 and Visio is bundled in.
• You need 2,000+ integrations and a mature enterprise support tier.

Skip it (use DBeaver or DataGrip’s ERD) if:

• You already use a database IDE that has built-in ER diagram generation — the Smart Query advantage disappears if your tool already handles this natively.

---

Alternatives Worth Considering

• dbdiagram.io — code-first, DBML syntax, free for personal use. Better for developers who prefer defining schema in text, weaker for visual-first teams.
• Lucidchart — 2,248+ reviews, mature collaboration, general-purpose. No database-native import. Starts at $9/user/month [4].
• DrawSQL — database diagram tool with import support. Fewer integrations than ChartDB’s engine list.
• DBeaver (ERD feature) — free community edition includes ER diagram generation from live connections. Requires direct DB access, unlike ChartDB. Better for local development than team documentation.
• DataGrip (JetBrains) — paid IDE with ER diagrams built in. Better for teams already in the JetBrains ecosystem.
• Miro — 4.7/5 on GetApp [4], 100M+ users. General whiteboard, not database-specific. No schema import.
• pgAdmin (for PostgreSQL only) — free, includes a basic ERD tool. Limited to PostgreSQL, far less polished than ChartDB for documentation purposes.

The realistic shortlist for a self-hosted database documentation tool is ChartDB vs dbdiagram.io. Pick ChartDB if visual import from an existing database is the priority. Pick dbdiagram.io if you prefer code-as-schema and want to version-control your diagram definitions in DBML.

---

Bottom Line

ChartDB solves one problem well: getting from an existing production database to a shareable ER diagram in under a minute, without handing anyone your credentials. The Smart Query approach is genuinely clever and the tool’s UI is clean enough that non-engineers can read the output even if they didn’t generate it. For solo engineers or small teams who want schema documentation without a monthly Lucidchart bill, the self-hosted path is straightforward and costs nothing beyond a VPS.

The honest caveats: real-time collaboration and persistent storage require the cloud plan, AI needs your own API key on self-hosted, and AGPL-3.0 means you need legal sign-off before embedding it in a product. If the cloud plan’s $59/month Teams tier fits the use case, that’s a fair trade against Lucidchart’s per-user pricing for a team that actually works with databases. If those collaboration features aren’t needed, self-hosting is a clear win.

If the Docker setup is the blocker, upready.dev deploys self-hosted tools like this for clients as a one-time service — you own the infrastructure, the bill stops growing.

---

Sources

1. LogRocket Blog — “Mastering charts and database visualization with ChartDB”. https://blog.logrocket.com/mastering-charts-database-visualization-chartdb/
2. ChartDB Official Docs — “Cloud vs. Self-Hosted”. https://docs.chartdb.io/docs/cloud-vs-self-hosted
3. SoftwareAdvice IE — “ChartDB | Reviews, Pricing & Demos”. https://www.softwareadvice.ie/software/530025/ChartDB
4. GetApp UK — “ChartDB Reviews, Prices & Ratings”. https://www.getapp.co.uk/software/2082123/chartdb
5. April, Medium — “ChartDB: The No-Install, AI-Powered Open-Source Database Diagrams Ed” (Oct 27, 2025). https://medium.com/@april-4/chartdb-the-no-install-ai-powered-open-source-database-diagrams-ed-78a4d351b379

Primary sources:

• GitHub repository and README: https://github.com/chartdb/chartdb (21,532 stars, AGPL-3.0)
• Official website: https://chartdb.io
• Live demo: https://app.chartdb.io