Teable 3.0: AI Agent Access, Pricing, and Self-Hosted Airtable Alternative Explained
- 2 hours ago
- 4 min read
Teable is an open-source, PostgreSQL-backed spreadsheet-database hybrid — an Airtable-style grid interface sitting on a real SQL backend, with AI fields, natural-language automations, and an app builder layered on top. Version 3.0's headline addition is the Teable Skill: external AI agents, including Claude Code and ChatGPT, can now connect directly to a Teable Base and manage its tables, fields, views, records, automations, and apps, rather than going through a generic Web API a developer has to wire up themselves.
The project runs under the AGPL-3.0 license with more than 21,000 GitHub stars, and ships both a cloud offering and a self-hosted edition with unlimited rows and storage. For a team evaluating it, the deciding factor is whether direct AI-agent access to the database itself — not just AI features inside the product — justifies choosing Teable over Airtable's proprietary platform or the other open-source contenders, given that Teable's cloud pricing still runs per seat.
··········
WHAT CHANGED IN 3.0, AND WHAT THE STACK IS BUILT ON.
A PostgreSQL backend, an AGPL license, and agent-native access define the technical shape of the product.
Teable's backend is a real PostgreSQL database rather than a proprietary storage layer, so data self-hosted through Teable is queryable and exportable the way any SQL database is — a structural difference from Airtable, where the underlying storage is closed. The AGPL-3.0 license on the community edition means the code can be run, modified, and even embedded in a product a team sells, a stronger guarantee than NocoDB's move to a Sustainable Use License in January 2026, which permits internal use but restricts resale.
The 3.0-specific addition is the Teable Skill: it exposes the base's structure and operations to AI agents in a form Claude Code or ChatGPT can act on directly — creating fields, modifying views, running automations, or building app pages — rather than requiring a developer to translate every agent request into individual API calls. App Builder in the same release added direct element selection (clicking a canvas element to attach it to chat for targeted edits) and inline static text editing, both aimed at shortening the loop between describing a change and seeing it applied.
........
Component | Detail | Function |
|---|---|---|
Backend | PostgreSQL | Data stays in a queryable, exportable SQL database |
License | AGPL-3.0 (community edition) | Self-hostable and embeddable, including in resold products |
Agent access | Teable Skill (new in 3.0) | Lets Claude Code, ChatGPT, and similar agents manage the base directly |
App builder | Direct element selection, inline static text editing | Shortens the loop between a described change and its application |
AI usage | Credit-based (per seat on cloud) | Meters AI chat, AI fields, and automation runs |
Self-hosted AI | Bring your own compatible language model | Keeps AI processing on infrastructure the team controls |
........
··········
WHY AGENT ACCESS IS A DIFFERENT CAPABILITY THAN AI FEATURES INSIDE THE APP.
Most competitors add AI to the interface; Teable Skill lets an external agent operate the base's structure itself.
AI fields and AI chat — auto-tagging records, summarizing content, answering natural-language queries about the data — are features a person triggers from inside Teable's own interface, and several competitors offer some version of the same idea. The Teable Skill is a different capability: it's an access surface for agents running outside the product entirely. A developer working in Claude Code can ask it to restructure a table, add an automation, or generate a new view without leaving their coding session, because the agent has a defined way to act on the base rather than needing a human to relay each change through the UI.
That capability changes what self-hosting buys a team beyond data residency. A self-hosted deployment can connect to a compatible language model of the team's choosing for AI features, which matters for teams that need AI processing to stay off third-party infrastructure — but the Teable Skill's agent-management capability is a product feature independent of hosting choice, available to any Claude Code or ChatGPT session pointed at a Teable Base, cloud or self-hosted. The trade-off is that AI usage — whether from in-app AI fields or agent-driven changes through the Skill — draws from the same credit pool, so a team leaning heavily on agent-managed automation should budget credits accordingly rather than assuming Skill access is free simply because it isn't a named line item.
··········
WHAT TEABLE COSTS AGAINST AIRTABLE AND THE OTHER OPEN-SOURCE CONTENDERS.
Per-seat pricing on the cloud plans still resembles Airtable's model; the self-hosted option is where the structural difference shows up.
Teable's cloud pricing runs Free, Pro at $13/seat/month (billed yearly), and Business at $20/seat/month (billed yearly) — materially below Airtable's $20 Team and $45 Business per-seat rates, but still a per-seat model that scales with team size the way Airtable's does. The self-hosted edition removes that scaling entirely: unlimited rows, unlimited attachments, and no per-seat charge beyond infrastructure cost.
........
Tool | License / self-host | Starting cloud price |
|---|---|---|
Teable | AGPL-3.0; self-hosted free (unlimited rows) | Free; Pro $13/seat/mo; Business $20/seat/mo (annual) |
Airtable | Proprietary, no self-host | Team $20/user/mo; Business $45/user/mo (annual) |
Baserow | MIT (core); self-hosted free | Cloud from $0 to $25/mo |
NocoDB | Sustainable Use License (since Jan 2026); self-host permitted, resale restricted | Cloud tiered, record caps by plan |
........
The comparison that actually matters is licensing intent, not the sticker price: Teable and Baserow's core carry OSI-approved-style open licenses permitting resale or embedding, while NocoDB's post-2026 license blocks offering it as a paid service to others. For a team just running its own internal base, that distinction rarely bites — but for anyone building a product on top of one of these tools, it decides which options are even legally available before price enters the conversation.
··········
THE DECISION RULE FOR EVALUATING TEABLE 3.0.
Teable earns consideration over Airtable in proportion to how much a team values PostgreSQL-backed data ownership and the option to self-host without a row cap — advantages that exist independent of the 3.0 release. The Teable Skill specifically is worth weighing separately: it matters to teams that already work through AI coding agents like Claude Code and want that agent to manage database structure directly, not just query it, and it matters less to teams whose only AI touchpoint is the in-app chat or auto-fill fields most competitors also offer. Teams planning to embed or resell a database layer inside their own product should treat the licensing table as the first filter, not the price column — Teable's AGPL-3.0 and Baserow's MIT core both permit that use, while NocoDB's Sustainable Use License does not, and no per-seat price difference changes that constraint.
·····
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
