Skip to content
callsend

Getting started

From nothing to your agents calling you.

  1. 01

    Hear your first call

    Sign in with GitHub and callsend rings you with a spoken brief on your repos, open PRs, and CI. Read-only access, about 60 seconds, no card.

  2. 02

    Get rings on this device

    On Home, click Enable rings. Your device then rings even with the tab closed. On iOS, install the app to your home screen first (Share, then Add to Home Screen); Safari requires it for push.

  3. 03

    Let your agent call you (remote MCP)

    Recommended, and zero install: point your client at callsend's remote MCP endpoint. It runs no code on your machine; the first call opens an OAuth sign-in your client handles.

    claude code
    claude mcp add --transport http callsend https://api.callsend.dev/mcp
# then run /mcp to sign in
    codex cli
    codex mcp add callsend --url https://api.callsend.dev/mcp

    Cursor, VS Code, or any other client — config JSON:

    mcp.json
    {
  "mcpServers": {
    "callsend": { "url": "https://api.callsend.dev/mcp" }
  }
}

    Headless or CI? Create a connector token in Settings and pass it as a header instead of signing in:

    terminal
    claude mcp add --transport http callsend https://api.callsend.dev/mcp \
  --header "Authorization: Bearer <token>"

    Your agent gets a call_me tool (reason and context); callsend rings you and leaves a voicemail if you miss it. Pairing scopes it: an agent can only ever call its own human.

  4. 04

    Full connector (dispatch to your machine)

    Optional. The connector adds speak-to-dispatch — hang up and your instruction runs on your machine through your own agent — plus a local stdio MCP. Requires Python 3.11+ and uv or pipx. It's published on PyPI, so install it by name:

    terminal
    uv tool install callsend    # or: pipx install callsend

    Or let the installer script pick whichever you have:

    terminal
    curl -fsSL https://callsend.dev/install | sh

    On Windows, use PowerShell:

    powershell
    irm https://callsend.dev/install.ps1 | iex

    Then pair it (opens a one-time approve link), serve your projects, and register the stdio MCP:

    terminal
    callsend login
callsend connect --project myapp=~/code/myapp --agent claude-code
claude mcp add callsend -- callsend-mcp

    --agent codex runs Codex instead. Repeat --project for each project; callsend status shows the stored pairing.

The loop

callsend rings you

When work lands, or when you start a call yourself from Home.

You talk

The voice knows your repos, PRs, and CI, and admits what it does not know.

End and Send

Hang up with the dispatch toggle on and your instruction runs on your machine, through your agent, in your project.

It rings back

A spoken brief when the run lands. Failed runs ring too, with an honest report.

Things people say on calls

There are no magic words. Talk like you would to a colleague who just called with an update:

  • “What changed since this morning?”
  • “Read me the failing test first.”
  • “Ship it, but add a regression test for that bug.”
  • “Hold off. Call me back when CI is green.”
  • “Switch to the payments project. Anything waiting on me?”

Interrupt it mid-sentence whenever you like; it stops and listens, like a person.

Good to know

Mic privacy

Audio streams only during a call you picked up; hanging up ends the stream.

Your agents, your machine

callsend never holds your repo credentials. Dispatched work runs locally under the agent you already use.

GitHub access

The demo and briefs read repos, PRs, and CI through a read-only GitHub App you can revoke anytime.

Billing

Flat monthly plans with a generous usage bar; no per-minute meter, and we flag you before a bar fills. See pricing.

For agents

An LLM-readable product reference lives at /llms.txt (full version at /llms-full.txt). Point your agent at it.

Hear it on your own repos

60 seconds from sign-in to a call. No card.

Sign in with GitHub