Speak AI

It requires SPEAK_API_KEY and SPEAK_BASE_URL. The API key is sensitive, and mishandling in configuration, logs, or downstream calls could enable misuse of the Speak account or API quota. The materials do not show credential requests beyond its stated purpose.

It declares network access to api.speakai.co, which is consistent with its stated functions of searching recordings, summarizing meetings, creating clips, and automating workflows. This implies related audio, transcript, or summary content may be sent to Speak's service for processing. No evidence is shown of data egress to unknown or unrelated endpoints.

System checks indicate the tool can execute code or spawn processes, but the materials do not disclose extra high-privilege system actions, persistence mechanisms, or execution behavior unrelated to its stated purpose. For an MCP tool, this is a typical capability and should be constrained by the runtime environment.

Based on the description, the tool may access user-provided recordings, meeting content, and derived summaries or clips. However, the available materials do not specify the scope of local file read/write access, and there is no indication of clearly excessive data permissions. With no README, data-boundary transparency is only moderate.

The source is an official registry entry, with a public open-source repository updated within the last year, all of which are positive risk-reducing signals. While community stars are low and the license is unspecified, limiting auditability and reuse clarity, there is no sign of closed-source exfiltration, untrusted origin, or obvious poisoning indicators.