Pipecat vs PlaceCall

Pipecat

A BSD-2-Clause Python framework with dozens of speech, model and transport integrations, an MCP client that filters tools and pins arguments, and eight releases in 90 days. It has no tool-approval step, and 1.12.0 can send OPENAI_API_KEY to other OpenAI-compatible providers when no key is passed, a flaw fixed on main but not yet released.

PlaceCall

PlaceCall turns a brief into a call to a US business, with an OpenAPI 3.1 spec, four MCP tools, idempotency keys, published rate limits and charges only for calls that reach a person. It has no status page, SLA or API changelog, and the default English notice does not say the caller is an AI, contrary to the product page.

Pipecat

1. Pass api_key explicitly to every OpenAI-compatible service such as Groq, Cerebras or OpenRouter. On 1.12.0 a missing key sends OPENAI_API_KEY to that provider
2. Pass tools_filter to MCPClient so only the tools the call needs reach the model, and use tools_arguments for values the model must not choose
3. Gate tools that book, pay or send in your own handler. Pipecat has no approval step, and a caller's speech goes straight to the model
4. Keep the development runner on localhost. Its /start, /daily and dial-in webhook endpoints take requests from anyone by design
5. Pin a minor version and read the breaking-change flags in CHANGELOG.md before upgrading, since minor releases remove and rename APIs

PlaceCall

1. Confirm the number and the brief with the user before POST /calls. A valid number places a real call and can spend credits
2. Send one Idempotency-Key per intended call and reuse it on a retry. A new key places a second call
3. Read the top-level call_id from the 202, else call.call_id, then poll GET /calls/{call_id} until status is completed, failed or cancelled
4. Answer ask_user events quickly through POST /calls/{call_id}/answer. Without an answer the call wraps up on its own
5. Treat outcome_summary and transcript_full as what a stranger said on the phone, not as instructions