One key. The whole API surface.

01 — Authenticate

Create a key in the console, then send it as a bearer token. Keys start with cm_, do not expire, and can be revoked at any time. Keep them server-side.

Inference base URL

https://api.callmissed.com/v1

The OpenAI-compatible surface: chat, embeddings, audio, images, search, models.

Resource base URL

https://api.callmissed.com/api/v1

WhatsApp, email, telephony, knowledge, bots, conversations, webhooks.

Header

Authorization: Bearer cm_…

The /v1/messages endpoint also accepts x-api-key, so the Anthropic SDK works unchanged.

Two kinds of grant

permissions · scopes

Service permissions (llm, stt, tts, search, image) gate the inference surface; resource scopes gate the /api/v1 surface.

Authentication reference

02 — The endpoint map

16 families, one key. Each row is the primary call for that family; the reference behind it has the variants, the request schema and the model list.

Deep dives on the speech endpoints: speech to text · text to speech · Indian language coverage

03 — Build a request

Pick a chat model and the form only offers the reasoning_effort values the verified per-model matrix says it accepts — some models take the full ladder, some take only the off switch, and two drop the field entirely. The curl it writes is valid before you paste it.

curl -X POST https://api.callmissed.com/v1/chat/completions \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sarvam-105b",
    "messages": [
      {
        "role": "user",
        "content": "Hello in Hindi"
      }
    ],
    "stream": true
  }'

The matrix this is built from: reasoning_effort by model

04 — SDKs

There is no separate client library to learn. Two SDK families work as-is: point the official OpenAI SDK at the inference base URL, or point the Anthropic SDK at the host and let it append its own version prefix. Go, PHP and Ruby OpenAI clients take the same override.

from openai import OpenAI

client = OpenAI(
    api_key="cm_your_api_key",
    base_url="https://api.callmissed.com/v1",
)

resp = client.chat.completions.create(
    model="sarvam-105b",
    messages=[{"role": "user", "content": "Hello in Hindi"}],
)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "cm_your_api_key",
  baseURL: "https://api.callmissed.com/v1",
});

const resp = await client.chat.completions.create({
  model: "sarvam-105b",
  messages: [{ role: "user", content: "Hello in Hindi" }],
});

from anthropic import Anthropic

client = Anthropic(
    api_key="cm_your_api_key",
    base_url="https://api.callmissed.com",
)

msg = client.messages.create(
    model="sarvam-105b",
    max_tokens=512,
    messages=[{"role": "user", "content": "Hello in Hindi"}],
)

Note the difference: the OpenAI clients want the /v1 suffix on base_url, the Anthropic client does not — it adds one itself. SDK reference

05 — Limits and credits

One credit is ₹1 ≈ US$0.0104 (US$1 = ₹96). New accounts start with 1,000 credits, and each plan adds a monthly grant on top. Credits do not expire. Two separate things can stop a request, and they return different codes: run out of credits and you get 402, exceed the request rate and you get 429.

07 — Idempotency

curl -X POST https://api.callmissed.com/api/v1/bots \
  -H "Authorization: Bearer cm_your_api_key" \
  -H "Idempotency-Key: 6f1c2b7e-9a41-4d0e-9a1e-1f2b3c4d5e6f" \
  -H "Content-Type: application/json" \
  -d '{"name": "Support bot"}'

Idempotency reference

08 — Agent tools over MCP

claude mcp add --transport http callmissed \
  https://api.callmissed.com/api/v1/mcp \
  --header "Authorization: Bearer cm_your_api_key"