rymi-live/rymi-mcp

List all your Rymi AI voice agents.

Retrieve a single Rymi voice agent by ID.

Fetch all available LLM models and voices you can use when creating or updating an agent. Always call this before create_agent to pick valid values.

List available agent/TTS voices, optionally filtered by provider or model. Much smaller payload than list_llm_options — use this to pick a `voice` value for create_agent/update_agent.

Preview the resolved per-language model stack (STT/LLM/TTS), blockers, warnings, model diffs, and any required role upgrades for a set of supported languages — without saving. Use before create/update to confirm a multi-language setup is valid. Note: the concierge (realtime) role is not supported he…

Create a new Rymi AI voice agent. Call list_llm_options first to discover valid voice and model values. For a multi-language agent, set supported_languages (e.g. ["hi-IN","en-US"]).

Update an existing Rymi voice agent's configuration. Only the fields you pass are changed; nullable fields accept null to clear them.

Permanently delete a Rymi voice agent.

Duplicate an existing agent. The copy gets " (Copy)" appended to its name.

List calls made with a specific agent.

Check whether an agent is ready to go live. Returns a validation report without persisting any changes.

Validate and resolve a flat key/value change-set against the AgentConfig field registry. Does NOT persist — follow up with update_agent.

Use AI + Google Search grounding to auto-generate a company description from a website URL, suitable for an agent's persona.

Use AI to generate an agent configuration draft from a plain-text description.

Estimate how much balance a call will consume for a given tier and duration. Results are usage estimates; surface remaining balance to customers in minutes.

Configure auto-recharge. `pack_usd` must exceed `threshold_usd` or the API rejects it (recharge-loop guard).

Configure spend-alert thresholds and low-balance / email preferences.

List previous and active calls across your account.

List calls currently in progress.

Retrieve details, participants, status, duration, and cost for a single call.

Retrieve the post-call summary for a call.

Retrieve the full transcript for a call. May contain personal data — handle accordingly.

Retrieve recording metadata (e.g. playback URL) for a call, when recording is enabled.

Retrieve current outbound call queue statistics.

Re-run post-call intelligence (summary, extraction, evaluation) for a call.

WARNING: places a real, billable PSTN call. `from_number` is the caller ID shown to the recipient and must be a phone number registered to your account — omit it to use the agent's attached number. `identity` is the destination in strict E.164 (e.g. +15555550123).

Queue up to 500 outbound PSTN recipients in one request. WARNING: places real phone calls and incurs per-minute charges. `from_number` is the caller ID shown to the recipient and must be a phone number registered to your account.

WARNING: immediately terminates an in-progress call. The call transitions to completed and post-call processing runs.

WARNING: dials/adds participants to a live call (warm transfer / conference). PSTN additions are billable.

List all numbers on your Do-Not-Call registry.

Check whether phone numbers are on the Do-Not-Call registry. Read-only — adds nothing.

Add a phone number to the Do-Not-Call registry so outbound calls to it are blocked.

Add up to 1000 numbers to the Do-Not-Call registry. Invalid numbers are skipped and returned in `invalid`.

WARNING: removes a number from the Do-Not-Call registry, re-enabling outbound calls to it.

Get this account's usage summary: remaining voice-runtime MINUTES, Studio AI unit usage, and post-call intelligence usage. Voice balance is reported in minutes, not dollars.

List published agent templates. Use a template's `defaults` as the starting config for create_agent.

List previous evaluation runs for an agent.

Retrieve a single evaluation run, including per-scenario scores.

Run the evaluation suite for an agent. mode="synthetic" (default) uses the offline scorer; mode="live" runs the model-driven runner (consumes Studio AI units).

List publishable (browser-safe) keys and which agent/channels each is scoped to. Returns key prefixes only, never full secrets.

List the knowledge sources (RAG context) attached to an agent.

List recorded configuration changes for an agent (for auditing or before an undo).

Add a knowledge source to an agent from raw text or a URL (the URL is fetched and ingested).

Delete a knowledge source from an agent.

Undo a single recorded configuration change, reverting that field to its previous value.

List all phone numbers on your Rymi account and which agent each is attached to.

Register a phone number on your account, optionally attaching it to an agent.

Attach an existing number to an agent so inbound calls route to it.

Remove a phone number from your account. The number stops routing to any agent.

Publish a Rymi voice agent to make its current saved configuration live. IMPORTANT: this immediately makes the agent callable by end users. Returns published:false with a blockers list when the config has unresolved issues.

Report whether a telephony carrier is connected, and which provider/account.

List numbers available on the connected telephony carrier account.

List your registered webhook endpoints. Signing secrets are never returned.

Register a webhook to receive call lifecycle events (e.g. call.completed, transcript.ready). URL must be public https. If you omit `secret`, one is generated and shown ONCE in the response — store it to verify signatures.

Update a webhook endpoint. Only provided fields change. Pass `secret` to rotate it.

WARNING: stops event delivery to this endpoint.