Airtable Alternatives: How to Choose and Plan a Safe Migration

Choosing an Airtable alternative starts with the job your base performs, not with a feature-counting exercise. A base with many-to-many links, rollups, formulas, attachments and API consumers needs a different replacement from a base that mainly tracks owners, statuses and deadlines.

Bending Spoons announced its Airtable acquisition on August 4, 2026, and completed it on September 4, 2026. That ownership change is a sensible reason to review dependencies. It is not, by itself, evidence that Airtable is shutting down, changing your plan or requiring a migration.

Before comparing products, document the base’s relationships, formulas, attachments, automations, interfaces, integrations, sensitive data and plan limits. A CSV can preserve exported values, but it does not preserve the working model or workflow that produced them.

The short answer: choose by what your Airtable base does

Use this decision gate before booking product demos:

• Relational data system: start with candidates that can represent linked records, relationship types and derived values. This is the right starting point when many-to-many links, rollups or downstream data consumers are load-bearing.
• Project tracker: test task-centered tools when the base mainly contains owners, statuses, priorities, due dates, comments and dependencies.
• Document-and-data workspace: test a document workspace when structured data is consumed alongside notes, specifications, meeting records or other prose.
• Temporary staging sheet: use a spreadsheet only as an interim cleaning or export location. It does not add relational integrity to the data.

Record the reason for switching before evaluating products. If the reason is only uncertainty about the acquisition, an inventory and plan review may be more appropriate than a migration.

Candidate	Start here when	Check before choosing
Baserow	You want to test a documented Airtable import path.	The importer requires a publicly shared Airtable base or view. Review the exposed scope and import report.
NocoDB	Self-hosting and deployment control matter.	Its current Sustainable Use License is source-available and fair-code, not OSI-approved open source. Review commercial-use terms.
Notion	Documents and structured databases need to coexist.	Relations work inside Notion, but CSV-exported relations become plain-text URLs and are not restored automatically.
ClickUp	The base is fundamentally task or project management.	Test relationship columns against the actual task model. Their existence does not establish Airtable parity.

These are candidates to test, not universal winners. Baserow documents an Airtable import process. NocoDB documents self-hosting and its current license. Notion describes its relations and rollups, while ClickUp documents relationship columns and supported calculations.

Inventory the base before comparing alternatives

Create one inventory row per table. Keep relationships, attachments, automations and exceptions in separate records because each has a different grain and validation method.

For each table, record the source table ID, record count, migration owner, relationship count, attachment count, formula count, lookup count, rollup count, automation count, interface count, downstream consumers, data sensitivity and proposed target owner. Mark which fields are authoritative inputs and which are derived display values. Save formula and rollup definitions before exporting.

For each relationship, record both endpoints and the relationship field ID. For each attachment, record the parent record, source reference, filename and verification status. For each automation, identify its trigger, action, downstream system and business owner. This prevents a table-level inventory from hiding a critical workflow.

Check Airtable’s current plan overview against the plan you actually need. As checked October 23, 2026, Free includes 1,000 records per base, cumulative across its tables, 1,000 API calls per workspace per month, 1 GB of attachment storage per base, 75 unique automations per base and 100 automation runs per workspace per month. Free also includes 500 AI credits per editor and above per month, unlimited read-only collaborators, five Editor or Creator collaborators and up to 50 Commenters.

Team is listed at $24 per collaborator monthly or $20 per collaborator monthly when billed annually. It includes 50,000 records per base, 100,000 API calls per workspace per month, 20 GB of attachment storage per base and 25,000 automation runs per workspace per month. Prices and entitlements can change, so confirm the applicable plan before making a decision. Airtable exports grid-view tables as CSV one table at a time. It does not provide one CSV for an entire base.

Prioritize bases with sensitive data, many-to-many relationships, business-critical automations or external API consumers. Archive unused bases instead of migrating them by default.

Shortlist by data model, not by feature count

Baserow: a documented importer, with configuration to rebuild

Baserow’s importer can transfer tables, records, field types, relationships, attachments and grid-view configuration. It also creates an import report for incomplete or incompatible items. Formula, Lookup and Rollup fields require reconfiguration. Automations, custom interfaces, integrations, row comments, revision history, field descriptions and access controls do not transfer as equivalent live configuration.

The importer requires a publicly shared Airtable base or view link, and authentication restrictions can prevent it from working. A shared view may expose only the data visible in that view. Do not make sensitive information public to simplify an import. Use an approved, sanitized test copy or another migration method.

NocoDB: separate deployment control from licensing

NocoDB is worth evaluating when internal self-hosting is important. Its documentation permits internal use and self-hosting, but its current Sustainable Use License is source-available and fair-code rather than OSI-approved open source. Offering NocoDB itself as a hosted or managed service to third parties requires a commercial license. Confirm the current license, deployment requirements and feature entitlements for the intended environment.

Notion: relationships inside a document workspace

Notion supports relations between databases and rollups across those relations. Its migration limitations are material: exported relation properties become plain-text URLs, CSV re-import does not restore working relations automatically, and a rollup cannot roll up another rollup. Test the actual relation structure and rollup logic rather than checking only whether a relation property can be created.

ClickUp: a task-centered option

ClickUp’s documented relationship columns can display fields from related tasks and support certain calculations, including sum, average or range where available. That documentation does not establish arbitrary multi-level rollups or full Airtable formula behavior. If ClickUp is shortlisted, test a real relationship, the required calculation and the user’s complete task flow.

For any candidate, run a proof of fit with one representative table, one relationship, one derived value, one attachment and one required user action. Check current plan entitlements against those requirements. Select the product that reproduces the necessary operation and access controls, not the one with the longest feature list.

Know what export and import actually preserve

Airtable’s table-level CSV export preserves values in a file. It does not preserve the formula, relationship, permission or workflow that generated each value. Treat exported identifiers and attachment references as migration inputs to validate, not proof that linked records or files have been recreated.

If using CSV, export each table separately and retain Airtable’s source record ID in a dedicated field. Build an approved mapping from source IDs to destination records before rebuilding links. In a duplicate source base, copy computed Formula, Lookup and Rollup values into static text or number fields if you need them for comparison. Preserve the original definitions separately and rebuild the logic in the target.

For Baserow’s importer, confirm that the shared link exposes exactly the intended scope. Import a sanitized test view before production data, include attachments only when appropriate, and inspect the import report. Verify transferred tables, records, relationships, attachments and views. Then configure unsupported formulas, lookups, rollups, automations, interfaces, integrations, comments, history and access controls as separate workstreams.

Baserow documents an option to update existing rows when imported data matches existing records. Use a stable source identifier and test the matching behavior. Do not treat that option as a concurrency-safe duplicate-prevention mechanism.

A controlled migration and cutover sequence

Use source IDs, not display names, as migration identity. An illustrative record key is source_system + source_base_id + source_table_id + source_record_id. For a relationship record, include both endpoints and the source relationship field ID. These are proposed implementation patterns, not fields required by Airtable or Baserow.

Keep the data grains separate:

• One record-level row per source record in one source table.
• One relationship-level row per link, especially for many-to-many relationships.
• One attachment-level row per file, including its source reference and destination file identifier.
• One field-validation row per record and field comparison.
• One migration-run row per attempt or batch.
• One exception row per failed record, field, attachment or relationship.

For repeated or concurrent imports, a read-then-create check can race. Two workers can both see no matching destination row and create duplicates. Prefer a destination-enforced unique key, a documented atomic upsert, a transactional insert-on-conflict operation or a queue with single-writer guarantees. If the target does not document atomic behavior, test retries and simultaneous runs and describe the control as an integration design, not a vendor guarantee.

Validate attachments at attachment grain. Confirm that every expected file exists, can be opened and is associated with the correct destination record. Copying a URL is not the same as transferring a file because URLs may expire or require source authentication.

Where AI can help, and where it should not decide

AI can assist with a bounded preparatory task such as suggesting categories for field descriptions or flagging likely formula dependencies in an exported inventory. It should produce a review queue, not a migration command.

A useful proposed output can include the source table ID, source field ID, suggested category, rationale and review status:

{
  "source_table_id": "tblOrders",
  "field_id": "fldRenewalDate",
  "proposed_category": "lifecycle_date",
  "rationale": "Description identifies a contract end date.",
  "review_status": "requires_human_approval"
}

This is an illustrative editorial design, not a documented Airtable, Baserow or ConsultEvo integration. Validate every returned table and field ID against the inventory, reject unknown identifiers and require a data owner to approve each classification before it affects the target schema.

Do not let a model decide relationship cardinality, authoritative values, access policy or whether an automation is safe to retire. Use deterministic rules for exact checks such as record counts, required fields, allowed values and relationship endpoints. Keep sensitive records and attachments out of AI workflows unless data handling, access and approval requirements have been reviewed. Airtable explains that AI-credit use varies by action and input in its AI billing documentation.

Make the shortlist decision and set a cutover gate

Choose database behavior first for linked-data workloads, task hierarchy first for project tracking and document usability first when prose and structured data belong together. Migration effort depends on relationships, attachments, formulas, integrations, access controls and the length of the parallel run. There is no reliable universal duration.

Do not cut over until the technical owner and business owner agree that the replacement passes these checks:

If you have selected ClickUp and want an implementation review against a task-centered workflow, see the ClickUp audit. The product choice remains yours. Migrate only when the target has passed both technical reconciliation and business acceptance.