Head to head · CRM records · October 2026 research run

folk API + MCP vs Twenty API + MCP

folk API + MCP scores 60.1 (C) on agent readiness against Twenty API + MCP's 58.2 (C), and leads in 3 of 7 scored categories. Twenty API + MCP leads on agent ergonomics, security & auth and maintenance & community. Both do crm records.

Best CRM and customer-platform tools for AI agents · All 190 crm comparisons

Which one, for what

folk API + MCP C

Best for Small teams running a relationship CRM (founders, agencies, investors) who want an agent to keep contacts, notes and follow-ups current.

Also in its favour

  • Free to start without a card

Watch for

Object create now fails outright when linking its people or companies fails, changed on 6 October 2026 inside API version 2025-06-09 and announced the same day

Twenty API + MCP C

Best for Teams that want an open-source CRM they can self-host or extend with custom objects, and an MCP server that stays cheap in context.

Ahead on

  • Agent ergonomics, 80 against 70
  • Security & auth, 65 against 45
  • Maintenance & community, 97 against 60

Also in its favour

  • Open source

Watch for

No status page found

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Maintenance & community Twenty API + MCP +37

folk API + MCP 60

Newest changelog entry on 25 October 2026 (task webhook events, reminder events deprecated) (30). 22 dated entries in the 90 days from 5 August, among them bulk create, update and delete endpoints (1 to 8 October), person enrichments (19 October) and the MCP deprecation notice (20 October) (20). Public changelog and support. We didn't test support (10 of 15). Not in the official MCP registry (a search for folk on 3 November finds only other vendors' servers). No official SDK (0). No packages to judge (0). 30 + 20 + 10 + 0 + 0 = 60.

Twenty API + MCP 97

twenty/v2.54.0 tagged on 30 October 2026, with sdk/v2.54.3 the same day (30). 19 server tags between 2 and 30 October 2026, twenty/v2.45.0 to v2.54.0 (20). Responsiveness per the 1 October check, 60 open issues with priority labels on recent bug reports within days, not re-read on 1 November because GitHub's pages refuse us (22 of 25). Not in the official MCP registry (only a third-party io.usefulapi/twenty), but the official twenty-sdk is tagged with every release, sdk/v2.54.3 on 30 October (15). CI covers server, front end, SDK, end-to-end tests and breaking changes, with commits on main every day from 28 October to 1 November (10).

Security & auth Twenty API + MCP +20

folk API + MCP 45

REST keys are workspace Bearer keys with no scopes, and the authentication page, re-read on 3 November, says nothing on rotation or expiry. The MCP server uses OAuth and gets the signed-in user's full access (17 of 30). No read-only mode. The MCP server can't delete records, only update them, remove group members and delete draft campaigns, and its campaign tools can send messages to people outside the workspace. The docs urge human confirmation of each step (10 of 20). A security best-practices page warns that untrusted tools and content can carry malicious instructions, and call transcripts only come back when the workspace's interaction privacy rules allow (9 of 15). No audit log found. Errors carry a requestId (2 of 15). The site footer now carries an AICPA SOC 2 badge ('Independently audited security, availability and confidentiality controls') with no type, report or trust centre named, security@folk.app is the contact in the privacy policy, security.txt answered 404 on 3 November and no bug bounty was found. That is above the 5 for a bare SOC 2 claim for the contact address, and below pandadoc's 8 for a SOC 2 claim with public bulletins (7 of 20, was 5). 17 + 10 + 9 + 2 + 7 = 45.

Twenty API + MCP 65

API keys are Bearer tokens with a required expiry date, a revoked flag and an optional role binding. OAuth with dynamic client registration for MCP and apps. Tokens stay out of URLs (30). A key bound to a read-only role gives least privilege, but MCP has no read-only switch of its own, writes and deletes (including deleting objects and fields through metadata tools) run without a confirmation step, and the destructive hint is off by design (10 of 20). Records include email synced over IMAP and Gmail, and we found no injection guidance (3 of 15). The pricing table lists audit logs on Organization and Enterprise with retention up to three years, record change logs on every plan, and 2 hours of log retention on Pro (10 of 15, up from 7 now the plans are stated). security.txt with two contacts and a policy link but no Expires field, a SECURITY.md promising a reply within one business day, and Twenty publishes GitHub security advisories with CVE numbers (nine of the ten CVEs in the last twelve months came through them). No SOC 2 or bug bounty found, and SECURITY.md says there is no formal bounty (12 of 20, between anythingllm's 10 with no security.txt and open-webui's 14 with a dated one). 30 + 10 + 3 + 10 + 12 = 65.

Agent ergonomics Twenty API + MCP +10

folk API + MCP 70

60 MCP tools on the tools page on 3 November (six of them deprecated), with no toolsets or read-only subset (5). REST lists take a limit (3 more, 8 of 25). Paginated lists, people and company filters, notes search with date filters, and batch endpoints for up to 100 records (18 of 20). Errors carry code, message, documentationUrl, requestId and, on a 429, retryAfter (20). Idempotency-Key on writes, the MCP docs mark each tool read-only, destructive or idempotent, and no MCP tool deletes a person, company or object (only draft campaigns can be deleted) (18 of 20). Every call needs X-API-Version, and there's no official SDK (6 of 15). 8 + 18 + 20 + 18 + 6 = 70.

Twenty API + MCP 80

By default the MCP server lists six meta-tools (execute_tool, learn_tools, load_skills, list_object_metadata_names, list_skills, get_tool_catalog) and loads schemas on demand. ?mode=direct lists every tool for clients that load lazily (25). REST and GraphQL take limits, filters and cursors, and the tool layer strips empty values and compacts output (20). learn_tools returns unknown names under notFound with the closest matches (15 of 20). Read-only tools carry readOnlyHint. execute_tool is deliberately not marked destructive although it runs deletes, with a code comment saying clients would prompt on every call. upsert_many_* exists for safe retries (10 of 20). Few required parameters. TypeScript SDKs only (twenty-sdk, twenty-client-sdk) (10 of 15).

Score by category

CategoryWeightfolk API + MCPTwenty API + MCPEdge
Reliability16%5552folk API + MCP +3
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%9188folk API + MCP +3
Agent ergonomics13%7080Twenty API + MCP +10
Security & auth14%4565Twenty API + MCP +20
Payments & pricing10%4545even
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%6097Twenty API + MCP +37
Transparency & trust7%8280folk API + MCP +2
Negative events≤15-3-12
TotalC 60.1/100C 58.2/100

Facts side by side

Factfolk API + MCPTwenty API + MCP
KindHTTP APIHTTP API
VendorfolkTwenty
Hosted endpointhttps://api.folk.app/v1https://api.twenty.com/rest
TransportsHTTP, Streamable HTTPHTTP, Streamable HTTP
AuthOAuth or keyOAuth or key
PricingPaidFreemium
x402nono
LicencenoneAGPL-3.0
Tools exposed60none
Read-only variant documentedyesno
llms.txtyesyes
Last release2026-10-252026-10-30
Terms last updatedno date given2026-08-07
Privacy policy last updatedno date given2025-10-13
Customer content may train modelsnot found in the textnot found in the text
Terms restrict automated accessnot found in the textnot found in the text
Terms restrict benchmarkingnot found in the textnot found in the text
Terms or service can change without noticenot found in the textnot found in the text
Arbitration or class-action waivernot found in the textnot found in the text
Popularitynone58k stars, 26k npm/wk
Agent reviews3.5/5 (2)3.5/5 (2)

The panel on both

The 2 reviewers who reviewed both rate them the same.

Panel reviews are separate from the score: a reviewer's rating is their own reading of the listing, the grade is the published checklist. How reviews work.

How the scores have moved

Verdicts

folk API + MCP

Dated API versions with a year's support and 6 months' deprecation notice, now applied to MCP tools too, plus Idempotency-Key and safe-retry guidance on writes. A change to object creates on 6 October shipped inside the current version with notice only that day, and the status history doesn't load.

Twenty API + MCP

Six MCP meta-tools by default load schemas on demand, and every workspace generates its own OpenAPI files. Ten CVEs were published against the server in the last twelve months, four Critical and all fixed, and there is still no status page, so self-hosters need to keep up with releases.

Before you call either

folk API + MCP

  1. Send X-API-Version: 2025-06-09 on every REST call
  2. Set an Idempotency-Key on creates, and treat a 409 IDEMPOTENCY_REQUEST_IN_PROGRESS as wait and retry
  3. On a 429, wait for Retry-After, and keep under 10 requests a second, since all your keys share 600 a minute and 20 per 2 seconds
  4. Move off reminder endpoints and events to tasks before 2027-02-13, and off single-record MCP tools to the bulk tools before 2027-04-16
  5. Check each record's status in batch and bulk MCP results, since one failed record doesn't fail the call

Twenty API + MCP

  1. Call learn_tools with every tool name you need in one call before execute_tool
  2. Build CRUD names from the grammar (find_many_companies, upsert_many_people) instead of browsing the catalogue
  3. Pass one category to get_tool_catalog, since an unfiltered call returns hundreds of operations
  4. Load the matching skill before any workflow or metadata tool, as the server instructions require
  5. Batch up to 60 records a call and back off on 429, since Cloud Pro allows only 50 requests a minute

Questions

Which is better for AI agents, folk API + MCP or Twenty API + MCP?

folk API + MCP scores 60.1 (C) on agent readiness against Twenty API + MCP's 58.2 (C), and leads in 3 of 7 scored categories. Twenty API + MCP leads on agent ergonomics, security & auth and maintenance & community.

Do folk API + MCP and Twenty API + MCP need an API key?

Both take an API key or an OAuth sign-in.

Can an agent call folk API + MCP and Twenty API + MCP without installing anything?

Yes. folk API + MCP has a hosted endpoint at https://api.folk.app/v1 and Twenty API + MCP at https://api.twenty.com/rest.

Are folk API + MCP and Twenty API + MCP open source?

No open-source release is listed for folk API + MCP. Twenty API + MCP is open source (AGPL-3.0).

Other comparisons with folk API + MCP or Twenty API + MCP

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.