clawcall
Server Details
Real US phone calls for AI agents: navigates phone menus, waits on hold, returns transcripts.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- OAuth
- Works in Glama
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- Repository
- ClawCall-Dev/ClawCall
- GitHub Stars
- 10
- Server Listing
- ClawCall MCP Server
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 19 tools
The set has several overlapping tools: place_call, place_call_and_track, and place_call_and_wait differ mainly in polling/wait behavior; call_card_action duplicates hangup_call and reply_on_call; call_card_state overlaps get_call/get_call_transcript/get_recording. Descriptions clarify intended contexts, but an agent can easily choose the wrong tool for the same underlying action.
Names are consistently snake_case and mostly follow a verb_noun pattern such as get_call, list_calls, hangup_call, and reply_on_call. Minor deviations like call_card_action/call_card_state use a noun-first pattern, but overall the convention is predictable.
At 19 tools, the surface is heavy for a calling server and includes redundant variants plus card-specific tools. The count is not extreme, but several tools feel like near-duplicates rather than independently earned capabilities.
The surface covers the core call lifecycle well: placing, tracking, replying, hanging up, transcripts, recordings, settings, balance, and webhooks. Minor gaps exist around webhook management and clearer card-vs-API separation, but the main workflows are covered.
Available Tools
19 toolscall_card_actionCall card actionADestructiveInspect
Call card only: the person's own taps. end_call hangs up and answer_question answers the phone agent's question, exactly as hangup_call and reply_on_call do; ui records a card interaction.
Inputs: action (required), card_id (required), name, final, +4 more. Returns: object
call_card_stateCall card stateARead-onlyInspect
Call card only: the call's status, transcript, open questions, takeover and recording in one read, by call_id or by the number a place tool is still dialing.
Inputs: to, call_id, card_id. Returns: to, call_id, outcome, takeover, +10 more
connect_webhookConnect ClawCall call eventsADestructiveInspect
Send this account's call events (call.started for inbound calls, call.completed for every call) to an agent's own webhook trigger, such as a Grok Bot routine, then send a test event and return its result. get_calling_guide topic webhooks explains the setup. The user copies the trigger URL and key from their agent app and pastes them to you. Saving sets Authorization as the only custom header, and ClawCall emails the account owner whenever events start going to a new URL. If the account already sends events elsewhere, this fails with the current destination unless replace is true; confirm with the user first. Calling it again with the same URL and key re-tests the connection.
Inputs: url (required), authorization (required), replace. Returns: test, webhook
get_balanceGet ClawCall plan and balanceARead-onlyInspect
Read the signed-in user's current plan, leftover prepaid balance, or trial allowance. Use after a quota/plan error or when the user asks what they have left. Explain allowance limits neutrally and preserve returned informational, sign-in, or support links.
Inputs: no named parameters. Returns: plan, trial, legacy
get_callGet call statusARead-onlyInspect
Read the current state of a call you own. Lifecycle is queued → dialing → answered → finalized. Returns a short transcript tail only. When finalized, read outcome and then get_call_transcript to judge the user's goal. outcome is the phone network, not task success.
Inputs: call_id (required). Returns: object
get_calling_guideGet ClawCall calling guideARead-onlyInspect
Use this when building a ClawCall call request or explaining how ClawCall works. Returns a markdown guide maintained with the ClawCall server: outbound (request fields and call workflow), examples (complete requests), errors, profile, inbound, privacy, or webhooks. call_format structured returns guides for the current call fields.
Inputs: topic (required), call_format (required). Returns: markdown
get_call_settingsGet ClawCall call settingsARead-onlyInspect
Read the signed-in user's voice and legacy/inbound personality, reserved-number state, and inbound-assistant availability, greeting, instructions, and knowledge base. Explains unavailable capabilities through entitlement and number status, with an informational account-access link.
Inputs: no named parameters. Returns: inbound, outbound, warm_greeting, reserved_number
get_call_transcriptGet full call transcriptARead-onlyInspect
Full turn-by-turn transcript for a call you own. Use this after lifecycle is finalized to decide if the user's goal succeeded. Do not treat outcome answered as success.
Inputs: call_id (required). Returns: transcript
get_recordingGet call recordingARead-onlyInspect
Temporary playable recording URL for a call you own. The URL expires; do not treat it as a permanent link. Offer it after summarizing the result; do not lead with the recording.
Inputs: call_id (required). Returns: recording_url
hangup_callHang up phone callADestructiveIdempotentInspect
Cancel or hang up a call you own. Idempotent; may return leg_stop_pending when carrier hang-up has not completed, including for finalized calls with outstanding cleanup. Retry the same call id.
Inputs: call_id (required). Returns: status, call_id, message, success
list_callsList call historyARead-onlyInspect
List compact call history for the signed-in user, without transcripts or call instructions. Use a returned id with get_call_transcript to read the conversation before judging the user's goal. direction=inbound requires Unlimited Reserve Plus and an active reserved number, or a number on the user's own Twilio account connected to ClawCall.
Inputs: limit, since, direction. Returns: calls, recordingWindowMinutes
open_call_audioOpen call audioBDestructiveInspect
Call card only: a single-use WebSocket URL for the call's live audio. talk takes the call over from the browser.
Inputs: mode (required), call_id (required), card_id. Returns: url, expires_in
place_callPlace phone callADestructiveInspect
Place one real outbound US phone call from a structured request. Returns call_id immediately; call track_call with it until the event is ended, then read get_call_transcript to judge whether the user's goal succeeded. outcome is the phone network, not task success. With delegation poll, the phone agent can ask you questions during the call, and you answer them through track_call. get_calling_guide topic outbound documents the request fields. The user may see a live card for the call.
Inputs: to (required), goal (required), authority (required), recipient (required), +9 more. Returns: hint, call_id, message, workflow, +1 more
place_call_and_trackPlace call and track itADestructiveInspect
Place one real outbound US phone call from a structured request, then wait for its first event, exactly like track_call. Keep calling track_call with the returned call_id until the event is ended. With delegation poll, the phone agent can ask you questions during the call; answer each with track_call's reply. get_calling_guide topic outbound documents the request fields. The user may see a live card for the call.
Inputs: to (required), goal (required), authority (required), recipient (required), +9 more. Returns: next, event, call_id, pending, +2 more
place_call_and_waitPlace call and wait brieflyADestructiveInspect
Place one outbound US call from a structured request and wait up to max_wait_seconds (default 45, max 90) for completion. Suited to a single call, not parallel campaigns. If the call is still running when the wait ends, call track_call with the returned call_id until the event is ended. get_calling_guide topic outbound documents the request fields. The user may see a live card for the call.
Inputs: to (required), goal (required), authority (required), recipient (required), +9 more. Returns: object
reply_on_callReply on a live callADestructiveIdempotentInspect
Answer a call.delegation event while its call is live. The phone agent passes your text on in its own words, usually in a sentence or two, and keeps the rest to share if asked. If the answer will take more than about 20 seconds, first send a short progress reply with final false, then the answer; after 45 seconds with nothing from you, the phone agent carries on without you. Another reply to an answered delegation adds to the answer, and an exact copy of an earlier reply is not passed on again. delivered false means the call has ended or the delegation is unknown, so tell the user directly instead.
Inputs: text (required), call_id (required), delegation_id (required), final. Returns: delivered, reply_result
take_over_callConnect the user to a callADestructiveInspect
Call card only: rings the account's loop-in contacts one at a time, or its verified phone when none are set, and connects the first phone that answers, voicemail included. The phone agent first tells the other side it is connecting them, then stops for good once a phone answers.
Inputs: call_id (required), card_id. Returns: call_id, message, takeover
track_callTrack a callADestructiveInspect
Wait for the next event on a call you own and return it: status when the call starts ringing or is answered, request when the phone agent needs you to look something up, waiting when nothing happened in this wait, superseded when a newer track_call on the same call took over, taken_over when a person took the call over from the phone agent, and ended when the call is over. Returns as soon as something happens. After status or waiting, call it again right away. Answer a request by calling track_call again with reply { request_id, text }; the phone agent passes the answer on in its own words. A reply with final false is a progress update: it keeps the question open and returns at once so you can finish the answer. reply_result accepted means the reply reached the call, not that it was heard yet. Another reply to an answered request adds to the answer, and an exact copy of an earlier reply is not passed on again. A request carries the whole call transcript so far, including what the other side said, so share only what the user's goal and authority need. Someone is waiting on the line, so answer quickly. When the event is ended, stop, read get_call_transcript, and tell the user about any unanswered questions or replies that never reached the call.
Inputs: call_id (required), reply. Returns: next, event, call_id, pending, +2 more
update_call_settingsUpdate ClawCall call settingsADestructiveIdempotentInspect
Patch reusable voice/personality or inbound-assistant settings in separate calls. Voice applies to all calls; personality applies to legacy and inbound calls only. Reading, changing, or clearing inbound configuration requires Unlimited Reserve Plus and an active reserved number, or a number on the user's own Twilio account connected to ClawCall; entitlement errors explain why the setting is unavailable and include an informational account-access link. Returns an acknowledgement only; call get_call_settings afterward when read access is available.
Inputs: voice, inbound, personality, warm_greeting. Returns: updated
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Give your AI agent a phone: place calls, navigate IVRs, wait on hold, get structured answers.
Give your AI agent a phone. Place outbound calls to US businesses to ask, book, or confirm.
Let AI agents place real phone calls from your verified number, with transcripts and recordings.
AI voice agents that make real phone calls: one number or a short list, with transcripts and notes.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to place real phone calls, navigate IVR trees, and retrieve structured answers with transcripts and recordings.MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to place real phone calls to US businesses in plain English — booking, canceling or rescheduling tables and appointments, verifying details, checking stock and requesting quotes — with IVR navigation and hold handling done automatically. Returns a labeled outcome, full transcript and recording, and can find candidate businesses when no phone number is known.121MIT
- AlicenseAqualityBmaintenanceLet your AI agent make phone calls and join Zoom, Microsoft Teams and Google Meet meetings for you. Smitline talks in real time, can get through phone menus, stays inside your brief, and returns the outcome, action items and transcript. Connect the hosted version (beta) at https://app.smitline.com/mcp, or self-host it: it's open source. Works with Claude Code, Codex, Cursor and any MCP client.15160Apache 2.0
- AlicenseNot gradedqualityCmaintenanceEnables AI assistants to place real phone calls, engage in spoken conversations across 12 languages, and report the outcomes for tasks like booking appointments, answering questions, negotiating, or confirming orders.MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.