total-recall, by Strvmarv, is an MCP server and knowledge base that provides persistent long-term memory for AI coding assistants. The app stores project conventions, past bug fixes, and preferences so agents retain context across sessions. It combines multi-tiered memory, semantic and full-text search, and a built-in local Web UI for browsing. It favors curated durable memory rather than indiscriminate capture, keeping stored context relevant.

What tasks can you actually use it for?

The tool supplies a persistent context layer so TUI coding assistants can reference historical project data instead of re-onboarding each session. It organizes information into daily logs for transient notes and durable registers for long-term decisions, and exposes stored context to agents through the Model Context Protocol. Practical outcomes include surfacing past bug fixes during debugging and applying consistent project conventions to new code reviews.

• Daily logs for raw notes
• Durable registers for lasting rules

How reliable are the memories for guiding future sessions?

Reliability depends on the tool's curation model rather than indiscriminate capture. A promotion filter moves only high-value items into persistent storage, which reduces noisy entries that confuse agents. A write-protection mechanism prevents overwriting detailed entries with shorter summaries, helping preserve rich context over time. These design choices aim to keep long-term registers stable and reduce accidental loss of historical detail.

What inputs and environment does it require?

The server ships as a single NativeAOT binary that runs on developer machines and is also available via npm for Node.js environments, so it can run without external runtimes when deployed locally. Compatibility requires an MCP-capable client; listed integrations include TUI assistants such as Claude Code and Copilot CLI. Deployments depend on routing Model Context Protocol calls from the assistant to the local server.

How does it fit developer workflows and privacy requirements?

The tool adopts a local-first data model, storing memories in a local SQLite database for offline availability and machine-level control. The single-file distribution includes embedded fonts and assets to simplify rollout on desktops. Teams gain direct control over retention and export, but they also take on database governance and periodic maintenance to prevent stale or inconsistent long-term records.

A pragmatic choice for teams that accept local maintenance

The tool is a practical option for developers who need persistent, curated project context and who accept the administrative overhead of running and maintaining a local memory store. Teams that formalize curation policies and schedule periodic reviews extract the most value, since consistent governance keeps long-term memory relevant and minimizes drift in assistant behaviour.