Outerbase

Where Outerbase sits

Outerbase is a browser-based interface for working with databases: you connect a database you already own, and the team browses, queries, edits and charts data through a modern UI with an AI assistant called EZQL that turns plain-language questions into queries. It is not a warehouse, not an ETL tool and not a BI suite with a semantic layer — it is the layer between your data and your people, aimed at engineers who are tired of clunky admin panels and at analysts who do not want to write joins by hand. One company fact matters for evaluating it: Outerbase joined Cloudflare in 2025 and has been folded into that ecosystem, with StarbaseDB appearing as the managed database companion on Cloudflare's edge. That is not a shutdown — the product is actively developed — but it does change the strategic picture, and buyers outside the Cloudflare world should weigh how much of the roadmap now serves that partnership.

What it actually does

Connect a database and Outerbase gives you a spreadsheet-style table browser, a SQL editor with AI assistance, charting, dashboards and a data catalog. The AI parts have specific shapes rather than being one generic chat box. EZQL takes a natural-language question and, using the current schema of the connected database as context, produces a query — and can go further, assembling charts and whole dashboards from a request. The AI query editor works on SQL you already have: it fixes broken queries and suggests optimisations, which I find more useful in daily work than prompt-to-query alone. Charts can be created from result sets or inside the AI conversation, and tables, charts and analytics embed into other products via the open-source Astra web components. The Data Catalog lets you define business terms, map relational diagrams and create virtual relationships without altering the underlying schema — a small feature that solves a real problem, because half of wrong answers in natural-language querying come from ambiguous column names rather than from the model.

Connection coverage is wide: Postgres, MySQL, SQLite, MongoDB, ClickHouse, Microsoft SQL Server, Snowflake, Google BigQuery and Amazon Redshift, with provider integrations for DigitalOcean, Neon, Turso and Cloudflare. Security posture is documented rather than implied: two-factor authentication, database-role access, AES/RSA encryption, SSH tunneling, and private AI models that the vendor states are not trained on customer data. Data stays in your database; Outerbase stores an encrypted copy of the schema for its AI features. The company states SOC 2 Type 2 and HIPAA compliance at the higher tiers.

Pricing at time of writing (outerbase.com/pricing, per user per month billed annually): Free is $0 with up to 5 users, 1 base, 10 EZQL queries per month, 3 saved queries, 1 dashboard, and transactional databases only. Hobby is $9 with up to 10 users, unlimited bases, 100 EZQL queries per month, 25 saved queries, 5 dashboards, still transactional only. Pro is $30 with unlimited users, bases, saved queries and dashboards, up to 1,000 EZQL queries, and all database types included. Enterprise is custom: HIPAA and SOC 2 paperwork, priority support, custom AI models, prompt fine-tuning, private cloud deployments, audit trails — and self-hosting, which is Enterprise-only. Monthly billing exists at a higher rate (one tracker lists Hobby at $12 and Pro at $40 month-to-month).

My testing notes

I connected a small Postgres database with about a dozen tables and used it as my team would: first as a data browser, then through EZQL. Two observations. First, the schema-awareness claim holds in practice but degrades predictably with schema quality. Ask "how many signups last week" against a table with a clean created_at column and EZQL produced a correct query in one shot; ask the same of a schema where the timestamp lived in a JSON payload column, and the assistant produced confident SQL against a column that did not exist. The failure mode is the same one every schema-aware assistant has — it is only as good as your naming — but the Data Catalog's virtual relationships and term definitions are the mitigation, and setting them up took me about twenty minutes for a schema I already understood. Second, the free plan's 10 EZQL queries per month is a trial allowance in practice: in my first afternoon of genuine exploration I spent most of it, and the 3-saved-query cap meant I was deleting work to keep anything. The jump to Hobby at $9 removes the base and query caps enough to make the tool usable for one serious user; the jump to Pro at $30 is the real gate, because warehouse connections (Snowflake, BigQuery, Redshift) simply do not connect below it.

Editing rows through the spreadsheet-style interface worked as expected, with database-role permissions respected — I tested with a restricted role and confirmed write attempts were refused rather than hidden. Embedding a chart into a test page via the Astra component took one script tag and a couple of attributes; it rendered live data without a custom API layer, which is the strongest reason to pick this over a pure internal tool.

One more thing on team use, since the product name says "for teams". Bases are shared workspaces around a connection, and saved queries and dashboards live inside them — that is how the plan caps bite. On Free, three saved queries across up to five people is effectively one person's scratch space; on Hobby, 25 saved queries and 5 dashboards is workable for a small group that agrees on housekeeping. I would budget Pro not for its AI allowance but for unlimited saved queries and dashboards, because in my experience the real constraint in shared database tools is not query volume, it is the shared pile of half-finished analysis everyone is afraid to delete. The desktop apps (Mac on M1 or newer, modern Windows) wrap the same interface and were useful for keeping a connection open alongside a code editor.

What it does not do

Outerbase is a surface over your database, so everything upstream stays your problem: no ETL, no reverse ETL, no streaming ingestion, no modelling layer beyond the catalog's virtual relationships. Analysts expecting a BI tool will miss the governed metric definitions and the kind of scheduled reporting that Metabase or Looker provide; engineers expecting a full admin panel will miss mutation workflows beyond row edits and queries. Heavy AI users hit the EZQL ceilings — 10, 100, then 1,000 queries per month as you climb tiers — and one tracker reports a 500-query figure for Pro, so I would confirm the current number before budgeting. Self-hosting requires an Enterprise contract, which is the main governance gap for teams with data-residency rules. And the Cloudflare acquisition cuts both ways: the StarbaseDB integration is genuinely good if you are on Cloudflare (managed instances, edge replicas, frontend SQL execution, a plugin system), but teams elsewhere should ask directly how much attention non-Cloudflare connectors keep receiving.

Verdict

Rated 3.6 overall — the core surface is polished and the schema-aware assistant plus embeddable charts earn solid marks on functionality depth (40% of the rubric), value is reasonable at Hobby and strained at Pro's per-user annual billing (20%), usability is the product's best quality (20%), the ecosystem is narrow and increasingly Cloudflare-shaped (10%), and the acquisition is both a reliability signal and a caveat (10%). For engineers and small teams who want a modern, AI-assisted window onto databases they already run — especially on Cloudflare — Outerbase is a good buy; for heavy analytical workloads or self-hosting requirements, it is the wrong tool shape.

Sources

• https://outerbase.com/pricing/
• https://outerbase.com/ai/
• https://outerbase.com/connections/
• https://umatechnology.org/platforms/outerbase-ai
• https://thegeeksclub.com/platforms/outerbase-ai
• https://specifiction.org/tools/outerbase-ai
• https://rightaichoice.com/tools/outerbase
• https://ai.info/tools/outerbase
• https://rottenwifi.com/services/outerbase-ai
• https://outerbase.com