MESSAGE
WORKSMW/01

Sendblue

Sendblue review: which plan fits your messaging workflow?

A public-source Sendblue review for choosing a plan, scoping an evaluation and recording the outcome of a message.

Browse practical guides →
Write two scenarios before choosing a plan: a customer asks for help, and your application starts a conversation. This is a public-document review; we have not sent test messages.

2 minute read · Provider review

Who starts the conversation?

Prices checked November 6, 2026. AI Agent lists 1,000 inbound contacts per day; that is not a messages-per-second rate. [1]

Use an actual customer task when choosing: answering a support question and initiating a reminder need separate entries in your requirements.

    Editorial comparison: Who starts the conversation?
    PlanMonthly USDConversation startCallbacks
    Free Sandbox0; shared numberVerified contact texts firstNone
    AI Agent100 per dedicated lineContact texts firstIncluded
    EnterpriseQuoteOutbound scope quotedIncluded

    Sources: [1], [3], [2]

    What does free testing establish?

    Free Sandbox permits ten verified contacts and has no webhooks. Another recipient must first text the shared number before you send to them; the quickstart describes actual iMessage/SMS traffic, not simulated delivery. It cannot establish production callback behavior. [1][2]

    For an evaluation, write the expected result beside each scenario. Keep a local replay result separate from what a consenting recipient observes and what reaches your application.

      Sources: [1], [2]

      What belongs in the bill?

      Sendblue’s terms add a 3% credit-card fee, applicable taxes and USD 60 per line to port out while the subscription remains active. After it ends, numbers return to the carrier and cannot be ported. Fees follow the order or written agreement. [3]

      Put the payment method, line count and planned exit date beside the quote, including how you would leave. [3]

        Sources: [3]

        What should your application record?

        V2 message records expose message_handle, service, status and was_downgraded. Acceptance, sending and reported delivery are different states. Webhooks require a 2xx acknowledgment; the docs specify a 45-second response window and up to three retries after server errors. [4][5]

        Store transitions, not just one seen/unseen flag per message; a later callback may update the same message. Decide who handles an unresolved outcome after retries stop. [5]

          Sources: [4], [5]

          What does mark-read mean?

          Mark-read requests a best-effort receipt for an existing one-to-one conversation on a supporting line. Success accepts the request; it does not confirm that the receipt appeared. This is not evidence that someone read your outgoing message. SMS has no visible read status. [6]

          An automated receipt and a human reply should have different labels in your product. Which does your workflow require before closing the task? [6]

            Sources: [6]

            What would make the plan a fit?

            Enterprise scope does not waive the terms: explicit opt-in is required, and unsolicited messaging and bulk SMS marketing are prohibited. [3]

            Before purchase, confirm the allocated line, permitted first-message scenario and environment for evaluating callbacks. Keep the answer with your quote. [1][3]

            How will an operator see a request that needs attention, and who owns that queue?

            The linked Crypton comparison retains its September 28, 2026 source check; it is not part of this update.

              Sources: [1], [3]

              Check Sendblue directly ↗

              Source notes

              1. Sendblue: pricing
              2. Sendblue: recipient setup
              3. Sendblue: terms — Effective March 19, 2025.
              4. Sendblue: message records — API v2.
              5. Sendblue: webhooks
              6. Sendblue: read receipts