Back to Updates

Audio capture for Long-Term Memory

Product

The biggest capture and integration release yet — LTM Audio, custom templates, and a 39-tool MCP server.

5.0.3 adds audio to Long-Term Memory Engine, puts full template control in your hands, and gives AI agents their deepest access to your work history.

What you can do now that you couldn't before

Capture your meetings and calls

LTM Audio (preview) records microphone input and system audio simultaneously, extracts context, and folds it into your Long-Term Memory. Pair programming sessions, stand-ups, client calls, and webinars become searchable alongside everything else you've worked on. Pieces processes audio locally; no raw audio is stored.

Build your own one-click summaries

Custom summary templates let you define exactly what you want: choose a time range, scope by website, project, or application, set your preferred prompt, and save it. Next time, one click generates the same structured output — no reconfiguring parameters.

Give your AI agent your full work history

The Pieces MCP Server now exposes 39 tools — full-text search, vector search, temporal queries, batch retrieval, and annotation access. Cursor agents, Claude Code sessions, and Goose workflows can query your Long-Term Memory as naturally as you can.

Reconstruct billable hours for any window

Time Breakdown now accepts configurable time ranges — the last 24 hours, the last week, or any custom date range — so reconstructions match your actual billing cycle.

Read math the way it was meant to be read

LaTeX expressions now render natively in Pieces chat and summaries. No more raw LaTeX syntax cluttering your conversations.

Who benefits most

  • Anyone in meetings, pair programming sessions, or video calls: LTM Audio.
  • Teams with recurring reporting formats and professionals who bill clients: custom summary templates.
  • Developers using Cursor, Claude Code, Goose, or any MCP-compatible agent: the expanded 39-tool MCP server.
  • Attorneys, consultants, government contractors, and CPAs: Time Breakdown with custom ranges.
  • ML engineers, data scientists, researchers, and academics: native LaTeX rendering.