github.devotd.
ReleaseRAG & retrieval

tedhaley/pensieve v0.9.0 — Connected agents as generative model

tedhaley/pensieve v0.9.0 is a Mac release that lets connected agents serve as a generative model, leaving work notes, summaries and insights over MCP.

Repositorytedhaley/pensieve ↗

tedhaley/pensieve v0.9.0 is a Mac release that lets connected agents serve as a generative model, leaving work notes, summaries and insights over MCP.

GitHub description: One place for all your information, for context engineering: your files, repos, agent sessions and tickets, on your Mac, for you and your AI agents.

stars
★ 65
Forks
6
Language
Python
License
MIT
Latest release
v0.9.0
Read original source ↗View on GitHub ↗

The GitHub release v0.9.0 of tedhaley/pensieve was published on October 17, 2026. What changed in this version is that connected agents can now be used as a generative model. Pensieve collects files, code repositories, agent sessions and tickets in one place on the Mac, a shared pool of information that people and AI agents use together. The purpose of this repository is to gather scattered information in one place for context engineering.

The one that can serve as a generative model is an external agent. Agents attach over MCP (Model Context Protocol, an interface that connects agents with external tools), call pending_work and submit_work, and leave work notes, summaries and insights as their results. Several agent sessions are processed at the same time, and the active sessions show up in Settings. The Models tab in Settings holds three cards, for the embedding model, the generative model and the MCP server, and it automatically discovers MCP servers such as Claude Code and Codex.

The confirmable scope, however, is narrow. This release is the Mac version, the material does not say which embedding model and generative model are used, and there are no measured performance figures. There is no license notice either, and the repository has 65 GitHub stars. The structure where several agents leave notes at the same time is understandable, but on this material alone it is hard to judge what makes it better than other tools.