# MeetCap 1.0: Local Meeting Notes Your Coding Agent Can Use MeetCap's first public Mac release combines local transcription, Markdown meeting records and an agent skill, with separate privacy and licensing boundaries for each workflow. By Paras Madan | 2026-10-19 MeetCap 1.0 lets an existing agent such as Claude Code or Codex record a meeting, help write its minutes and search earlier conversations. Its first public release arrived on October 18 for Apple Silicon and Intel Macs running macOS 14.2 or later. Recordings are transcribed locally, and the meeting library keeps transcripts and minutes as Markdown files. The [release announcement](https://github.com/seacen/meetcap/releases/tag/v1.0.0) describes free recording without an account or time limit, with summaries written by the AI the user chooses. For developers, the interesting part is having meeting decisions available to the same agent that works with their project files. ## A meeting library inside the agent workflow MeetCap is a standalone application with a command-line interface and an accompanying agent skill. The skill teaches compatible agents how to operate the `meetcap` command, including starting and stopping recordings, transcribing an existing file and finding what someone said in a previous meeting. The [versioned skill](https://github.com/seacen/meetcap/blob/v1.0.0/skills/meetcap/SKILL.md) also covers summaries, minutes, speaker names, scheduling and document exports. It tells the agent to start or stop a recording only when the user asks. Destructive actions, such as deleting a meeting or forgetting a saved speaker, have a separate confirmation step. That makes the library useful for a question such as which requirements changed in the last project meeting. The agent can retrieve the recorded discussion alongside other project material, then use it as context for the next task. Before acting on a decision, the user still needs to check that the agent found the right meeting and interpreted it correctly. The skill also tells agents to treat transcripts and other meeting material as data. A sentence spoken during a meeting should not become an instruction to send a file, change permissions or perform an unrelated action. These are the project's documented rules; their presence is not an independent test of how every compatible agent behaves. ## Local transcription and selectable summaries The first release advertises English, Chinese and mixed English/Chinese transcription. It describes periodic live summaries, speaker identification and minutes whose points link back to transcript lines and audio playback. Those links provide a way to review a claim against the recording instead of relying entirely on the generated summary. For writing summaries and minutes, users can select an agent already installed on the computer, their own API provider or a local model that MeetCap downloads. The application itself is free without feature locks. A remote provider's normal fees or subscription allowance still applies when that provider does the writing. There is an operational detail in the agent skill worth noticing. When the chosen agent is responsible for writing minutes, MeetCap leaves pending jobs for it to process. The skill tells the agent to check those jobs after recording stops and when the user returns to the meeting. If that agent is unavailable, the job can remain waiting. For a first meeting, check the whole sequence: capture both sides, finish transcription, generate minutes and open a few of their source links. A transcript appearing in the library does not establish that every later step completed. ## What stays on the Mac? The [privacy notice for version 1.0](https://github.com/seacen/meetcap/blob/v1.0.0/PRIVACY.md) says recordings stay on the device that made them. Transcripts, minutes and speaker voiceprints live in the meeting-library location the user selects. When a remote AI writes the summaries, MeetCap sends it transcript text, meeting background and glossary context. An agent that can read the library can also handle its contents under that agent provider's terms. Choosing a local summary model keeps that text processing on the device, according to the project. Storage is another choice. Putting the library in iCloud Drive, OneDrive or a similar service brings that service's syncing and access rules into the workflow. A company evaluating MeetCap should review the summary provider and the library location separately, especially for meetings containing customer information or unpublished work. Speaker recognition adds a further consideration. Naming a speaker saves a voiceprint and supporting audio clips in the library so the person can be recognised in later meetings. The notice says it does not save voiceprints for unnamed speakers. Anyone with access to a synced library may also be able to access the saved voiceprints. MeetCap's terms require users to tell participants about recording and obtain applicable consent. They also address consent for saving voiceprints. Local processing still leaves those responsibilities with the person recording the meeting. ## Downloads, licensing and an initial trial The [1.0 release assets](https://github.com/seacen/meetcap/releases/tag/v1.0.0) include separate Mac downloads for Apple Silicon and Intel. Windows and iPhone versions are described as coming soon, with no release date established in the announcement. The public GitHub repository can be misleading if read as an open-source claim. MeetCap's [license](https://github.com/seacen/meetcap/blob/v1.0.0/LICENSE) describes a proprietary application with free personal and commercial use. The accompanying agent skill and reference files use MIT-0. The application source is not supplied under that license. This is a documentation-based overview, without a hands-on accuracy or speaker-recognition benchmark. A useful trial would use a short, non-sensitive meeting with a known decision and a few project-specific names. Check the transcript, inspect the minutes' supporting lines and ask the agent to find that decision later. That last retrieval is especially relevant to the product's promise. A meeting archive becomes useful project context when the agent can recover the right statement and show where it came from. Keeping the records as ordinary files gives the user a way to inspect them and keep using them outside MeetCap. I'm Paras. I work at the intersection of AI, engineering and distribution. I help AI and developer-tool teams turn product workflows, privacy choices and technical tradeoffs into clear stories for the people who can use them. [Book a 30-minute call](https://cal.com/paras-madan/30min).