Inworld Realtime API vs PlaceCall

Inworld Realtime API 62

Hosted reading of the checklist. status.inworld.ai is a public status page (20), with history we couldn't read, as on inworld-tts (5). Concurrent sessions are published per plan, 10 on On-Demand up to 1,000 on Growth (15). The rate limits page gives exponential backoff with jitter and Retry-After on longer windows, and sessions emit rate_limits.updated, with nothing on reconnecting a dropped session (12 of 15). An SLA appears only in the Enterprise column, with no terms (0). No preview or beta label on the Realtime API (10). 20 + 5 + 15 + 12 + 0 + 10 = 62.

PlaceCall 45

Read with the hosted lines. No status page was found on voygr.tech or in the repository, and status.voygr.tech still does not resolve (0). With no published history the incident line takes 5. The skill in the repository publishes rate limits of 10 requests a second and 120 a minute on free and paid keys, plus a daily call ceiling (15). It says to honour Retry-After on 429 and resume_at on a 503 for maintenance, and every POST /calls carries an Idempotency-Key whose replay returns the stored response with Idempotent-Replayed: true (15). No SLA for any tier (0). Nothing in the skill, README or product page calls the API beta or preview (10). api.voygr.tech's robots.txt now disallows ClaudeBot and names neither Claude-User nor Claude-SearchBot, so the 17 October readings of its OpenAPI spec, llms.txt, llms-full.txt, docs, checkout pages, /v1/pricing and protected-resource metadata are struck, and these lines now rest on the skill repository. 0 + 5 + 15 + 15 + 0 + 10 = 45.

Inworld Realtime API 68

Scored as an API. Conversation history defaults to the last 100 items, can be trimmed, and can be cached on Anthropic models, and every response.done carries a usage block (18 of 25). No lists to page. Output controls are modalities, turn detection settings, optional transcription and response.speak (12 of 20). Error events follow OpenAI's type and code shape, unsupported fields come back as non-fatal unsupported_field warnings, and item_not_found names the bad parameter (14 of 20). Tools run on the client and a follow-up response starts automatically after a tool result. Nothing covers side effects of a tool call cut off by an interruption (12 of 20). Works with OpenAI Realtime clients after changing the URL and key, with no Inworld Realtime SDK (12 of 15). Total 68.

PlaceCall 81

Four MCP tools per the README and the skill (suggest_places, place_call, get_call_result, answer_question), so context cost is small (25). GET /calls takes limit, lists most recent first without transcripts, and the event feed resumes from an after_event_id cursor, but no filters on the call list were found (12 of 20). Errors carry an error_code and a hint, a 422 for missing slots lists each one with a suggested_question, and a 402 names needed_credits (20). Idempotency keys on POST /calls with 409 idempotency_conflict, idempotency_in_progress and idempotency_state_unknown answers, and a retry procedure that lists calls before re-dialling. MCP tool annotations were not readable (16 of 20). A call needs only target_phone and brief, with language defaulting to auto. No SDK packages in any language, only plugins, a skill and the API (8 of 15). 25 + 12 + 20 + 16 + 8 = 81.

Inworld Realtime API 63

Closed service with clear terms naming Theai, Inc. dba Inworld AI (15). The ZDR page says what a session keeps and that third-party LLM requests aren't covered, but the terms and privacy notice disagree on training (12 of 30). Release notes flag potentially breaking changes with the clients affected and what to do, without stating a notice period (14 of 20). US headquarters with EU and India residency on Enterprise, and no sub-processor list in the docs, while sessions may route to third-party LLM providers (10 of 20). Total 51.

PlaceCall 50

Closed service with dated API terms (last updated 23 October 2026), and the skill and manifests are MIT (15). The API terms and the privacy policy (6 October 2026) agree on 90 days' retention for recordings and transcripts. A DPA is available only by negotiation, the privacy policy names VOYGR an independent controller of call recipients' data for its own purposes, and it allows de-identified data to be used and shared (20 of 30). The terms reserve the right to change or remove endpoints with or without notice and promise no backward compatibility. The skill marks outcome_type deprecated with no date, and the merged-transcript endpoints were removed with no dated notice (3 of 20). No subprocessor list or data locations. The privacy policy names only categories of service providers (5 of 20). 15 + 20 + 3 + 5 = 43.

Inworld Realtime API 40

No x402, MPP or L402 (0). No per-session price, but each billed part (STT per hour, TTS per character, LLM per token) is priced publicly (20). On-Demand includes free STT and TTS allowances and lists no payment method, though some premium LLMs need a card on file (20). A person signs up in a browser before a key exists (0). Total 40.

PlaceCall 50

No x402, MPP or L402. A 402 from POST /calls is a quota_exceeded error with needed_credits and a checkout_url, not a payment challenge (0). The product page publishes $0.10 per successful call, $0.096 or $0.09 on the 2.5K and 5K credit packs, and $0.05 per answered place suggestion (20). 250 free calls (2,500 credits) on a new account, and both the MCP sign-in and the email-code sign-up create it with no card step (20). A key is issued by POST /signup and a six-digit code sent to the user's email, so an agent with its own inbox can get one. The MCP route needs a person to sign in, and buying credits goes through Stripe checkout (10 of 20). 0 + 20 + 20 + 10 = 50.