Mira is a chat interface built around continuity across conversations. It keeps a single unbroken thread rather than starting over each time, and its memory is meant to grow from repeated use instead of being reset.
The system automatically extracts and curates memories from conversation. Items that are revisited keep their weight, while those left alone fade. Mira describes this memory system as being designed like neural pathways: overlapping memories can merge, and they can accumulate typed relationships such as corroborating, conflicting, and superseding. The page also says the system maintains itself and does not require human-in-the-loop oversight.
A feature called Domaindocs provides a shared notebook for project notes, references, and ongoing context. These notebooks are persistent, sectioned, and jointly written, and both the user and Mira edit the same document in real time. Sections can be collapsed so the notebook stays scannable. The page also says Mira has a self-model or adaptive intelligence that learns from how a user responds, including what lands, what backfires, and when pushback or acknowledgment works. It presents each Mira as paired with one person and accumulating user-specific skills over time.
Mira is available as a hosted service and as a freely available open-source repository. The entry does not name a specific pricing plan, and it does not identify a maker beyond the page’s own references to the developer.
In the Chatbots & assistants space, Mira takes a focused approach. It focuses on maintaining context and continuity in AI chat conversations without losing previous information. It is built as a consumer product for users seeking persistent AI chat assistants. Mira is open source under the AGPL-3.0 license. It ships for the web.
Mira first shipped in 2025. The project is developed in the open on GitHub with 467 stars and 36 commits in the last 90 days. Among its 5 catalogued features are persistent chat memory, context-aware responses, and shared notebooks.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do