Head to head · SMS · October 2026 research run

Linq vs Prelude

Linq scores 71.7 (BB) on agent readiness against Prelude's 67.6 (B), and leads in 4 of 7 scored categories. Prelude leads on security & auth. Both do sms.

Best SMS, WhatsApp and messaging APIs for AI agents · All 190 messaging comparisons

Which one, for what

Linq BB

Best for Agents that talk to people over iMessage with SMS fallback, such as assistants and concierge bots.

Ahead on

  • Reliability, 83 against 65
  • Agent ergonomics, 90 against 85
  • Payments & pricing, 40 against 30
  • Transparency & trust, 72 against 42

Also in its favour

  • Agent-ready, a grade of BB or better
  • Runs on your own machine
  • Free to start without a card

Watch for

Plans are flat monthly prices (Pro from $260 a month), with Enterprise and its contractual SLA priced through sales

Prelude B

Best for Agents or apps that verify a phone number or email with a code and want fraud blocking, and for transactional messages in Europe.

Ahead on

  • Security & auth, 63 against 50

Watch for

The terms, privacy policy and legal notice are Notion pages drawn by script, unreadable by our reader

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.

Transparency & trust Linq +30

Linq 72

Closed service with clear terms from Linq App Inc. (14 April 2026); SDKs, MCP server and CLI are Apache-2.0 (15 of 30). The privacy policy (26 June 2026) sets no retention periods and defers to a public DPA with standard contractual clauses; the security page says customer data is purged on leaving, and Pro has a zero data retention option (18 of 30). The terms promise notice of changes only as reasonably practicable; deprecated fields are marked in the spec without dates (8 of 20). Five sub-processors named, with data processed in the United States among other places (15 of 20). 15 + 18 + 8 + 15 = 56.

Prelude 42

Closed service whose general terms are a Notion page drawn by script, which we couldn't read; the SDKs are Apache-2.0 (8 of 30). The privacy policy is also unreadable for the same reason, and the trust centre names no retention periods or data locations (3 of 30). v1 and v2 are documented side by side and the changelog dates changes by month, with no deprecation policy found (6 of 20). No sub-processor list or data locations found (0 of 20). 8 + 3 + 6 + 0 = 17.

Reliability Linq +18

Linq 83

Read with the hosted lines. status.linqapp.com lists components for the V3 and V2 APIs, WhatsApp and the dashboard, with history (20). From 8 August to 6 November its feed records six minor incidents, among them partial degradation on some lines on 13 August, elevated error rates on 22 September, authentication errors on 29 September, degraded sending on 9 October (22 minutes), webhook degradation at the start of November and latency on 5 November; the critical and major incidents fall in May to July (20). Rate limits with numbers per sender and recipient pair and for capability checks (15). 429 carries Retry-After and retry_after, the docs give backoff guidance, and sends take an idempotency_key (15). llms.txt states 99.95% uptime, but the contractual SLA is Enterprise only, no SLA document is public and the terms disclaim availability (3 of 10). Partner API v3 is generally available (10). 20 + 20 + 15 + 15 + 3 + 10 = 83.

Prelude 65

Read with the hosted lines. status.prelude.so has components for Verify, Auth, Watch, Intel, Notify and the dashboard, with history (20). From 8 August to 6 November its feed records one major incident, increased error rates and latencies across the APIs on 16 and 17 August, posted after it was resolved with no duration stated, and minor ones such as Verify latency on 9 and 10 August, SMS degradation in China and on Orange France and delayed webhooks on 23 October (10). No numeric API rate limits were found; only the MCP server's limits and per-window Verify retry limits are documented (7 of 15). A 429 carries Retry-After and X-Should-Retry, and repeating the create call inside the window performs a retry rather than a new verification (13 of 15). The pricing page lists a 99.95 SLA on Enterprise with no SLA document found (5 of 10). Verify and Notify v2 are generally available (10). 20 + 10 + 7 + 13 + 5 + 10 = 65.

Security & auth Prelude +13

Linq 50

Bearer API keys in a header, never in the query string, rotated or revoked in the dashboard, scoped to the account with no per-action or read-only scopes (20 of 30). Sandbox keys reach only verified handles, and the MCP server's --code-allow-http-gets and method allow and block lists limit its code tool; no approval step for sends (10 of 20). Inbound iMessage and SMS text reaches the agent, and no prompt-injection guidance was found in the docs or the MCP instructions (0 of 15). Every response carries a trace_id; no operator request log was found in the docs (5 of 15). SOC 2 Type II reports for 2024, 2025 and 2026 and a 2026 penetration test on request, a vulnerability disclosure policy open on the trust centre and SECURITY.md in the SDK repository; no security.txt or bug bounty (15 of 20). 20 + 10 + 0 + 5 + 15 = 50.

Prelude 63

API keys as bearer tokens with no scopes found; the MCP server uses OAuth with a per-user consent screen (20 of 30). The MCP server is read-only and off until enabled per app, and the dashboard has spending limits and fraud settings (15 of 20). Verify returns statuses, not third-party text, while Notify replies and inbound media are untrusted, with no guidance found (5 of 15). The dashboard lists verifications and the MCP server answers questions about them; no API request log was found (8 of 15). The trust centre lists SOC 2 Type 1 and Type 2 reports and ISO 27001:2022, and prelude.so has a valid security.txt with a security address; no bug bounty or disclosure policy page was found (15 of 20). 20 + 15 + 5 + 8 + 15 = 63.

Payments & pricing Linq +10

Linq 40

No x402, MPP or L402 (0). Public plan prices only, Hobby $0 and Pro from $260 a month, with no per-message fee (10 of 20). Hobby is free with no card, and the Pro trial needs no card (20). linq signup creates an account, an API key and a number from a terminal, but a person has to read back the emailed code, so half (10 of 20). 0 + 10 + 20 + 10 = 40.

Prelude 30

No x402, MPP or L402 (0). Per-verification price (€0.032) and per-message Notify prices published without login (20). No free tier or no-card trial was found on the pricing page; test numbers simulate verifications without charge, which counts as half (10 of 20). A person signs up in the dashboard (0). 0 + 20 + 10 + 0 = 30.

Agent ergonomics Linq +5

Linq 90

The MCP server has two tools, code execution against the SDK and docs search, so its definitions cost little context (25). Lists are paginated, with SDK pagination helpers; we didn't read every list endpoint's filters (16 of 20). Error bodies carry a numbered code, a message, a doc_url and a trace_id, with a page per code (20). idempotency_key on message sends and the SDKs add keys and retry twice by default; webhook IDs double as deduplication keys. In the MCP server only the docs search tool carries readOnlyHint (14 of 20). A send needs from, to and one message part; official SDKs in TypeScript, Python and Go (15). 25 + 16 + 20 + 14 + 15 = 90.

Prelude 85

List calls take limit from 1 to 200 with cursor or offset; the MCP server's tool list isn't on the docs page, so its size is unchecked (18 of 25). Cursor and offset pagination with filters (18 of 20). Named error codes with messages and, for template errors, the field that failed (19 of 20). Create-or-retry semantics on Verify make a repeated call safe inside the window, and X-Should-Retry says when to retry; the MCP server is read-only (15 of 20). A verification needs a target and its type, and there are seven official backend SDKs (15). 18 + 18 + 19 + 15 + 15 = 85.

Score by category

CategoryWeight this runLinqPreludeEdge
Reliability16%208365Linq +18
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%16.29393even
Agent ergonomics13%16.29085Linq +5
Security & auth14%17.55063Prelude +13
Payments & pricing10%12.54030Linq +10
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8.88383even
Transparency & trust7%8.87242Linq +30
Negative events≤15-20
TotalBB 71.7/100B 67.6/100

Facts side by side

FactLinqPrelude
KindHTTP APIHTTP API
VendorLinq App Inc.DING SAS (Prelude)
Hosted endpointhttps://api.linqapp.com/api/partner/v3https://api.prelude.dev
TransportsHTTP, stdioHTTP, Streamable HTTP
AuthAPI keyOAuth or key
PricingPaidPay per use
x402nono
LicenceProprietary service under Linq App Inc.'s terms of service and sale. The TypeScript, Python and Go SDKs, the MCP server and the CLI are Apache-2.0Proprietary service; the terms are on a Notion page we couldn't read. The backend SDKs are Apache-2.0
Tools exposed2none
Read-only variant documentednono
llms.txtyesyes
MCP registryio.github.linq-team/linqnot listed
Last release2026-11-052026-10-16
Terms last updated2026-04-14couldn't be read
Privacy policy last updated2026-06-26couldn't be read
Customer content may train modelsnot found in the textcouldn't be read
Terms restrict automated accessnot found in the textcouldn't be read
Terms restrict benchmarkingyescouldn't be read
Terms or service can change without noticenot found in the textcouldn't be read
Arbitration or class-action waiveryescouldn't be read
Popularity264k npm/wk98k npm/wk

Verdicts

Linq

Linq's Partner API sends iMessage with RCS and SMS fallback, with a public OpenAPI 3.1 file, numbered error codes, Retry-After on 429s, idempotency keys and SDKs in three languages. Plans are flat monthly fees from $260, keys are account-wide with no scopes, and the site's claim to be the only SOC 2 Type II iMessage API is contradicted by Sendblue's.

Prelude

Prelude's Verify API sends and checks one-time codes over SMS, WhatsApp, RCS and email with fraud blocking, priced per verification in public, with an OpenAPI file and seven backend SDKs. Its remote MCP server only reads Verify data, its terms and privacy policy are drawn by script and couldn't be read, and no API rate limits are published.

Before you call either

Linq

  1. Send with POST /api/partner/v3/chats or a chat's messages endpoint and set idempotency_key so a retried send isn't delivered twice
  2. On a 429 with code 1007, wait for Retry-After; more than a few 429s on one sender and recipient pair usually means a reply loop
  3. Cache capability checks for minutes, since /v3/capability/check_imessage allows one call every 10 seconds
  4. Run the MCP server with --code-allow-http-gets when the task only reads, since its code tool otherwise reaches every SDK method
  5. Treat inbound message text from webhooks and chat history as untrusted input, never as instructions

Prelude

  1. Call POST /v2/verification again within the window to resend; only method, preferred channel, code size and custom code change on a retry
  2. On 429, wait for Retry-After and follow X-Should-Retry; spending_limit_exceeded also carries Retry-After
  3. Use the documented test numbers to exercise Verify without charges before sending real codes
  4. Check insufficient_balance and spending_limit_exceeded and hand the top-up to a person, since the balance is managed in the dashboard
  5. Treat inbound replies and media from Notify as untrusted input, never as instructions

Questions

Which is better for AI agents, Linq or Prelude?

Linq scores 71.7 (BB) on agent readiness against Prelude's 67.6 (B), and leads in 4 of 7 scored categories. Prelude leads on security & auth.

Do Linq and Prelude need an API key?

Linq needs an API key. Prelude takes an API key or an OAuth sign-in.

Can an agent call Linq and Prelude without installing anything?

Yes. Linq has a hosted endpoint at https://api.linqapp.com/api/partner/v3 and Prelude at https://api.prelude.dev.

Other comparisons with Linq or Prelude

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.