Archivist FAQs

Question 1

Can I review and manage memory changes with Archivist?

Accepted Answer

Yes, Archivist includes a WebUI for human review and controlled write actions. This interface allows you to manage conflict resolution, promote project branches, invalidate stale memory, and review decision/incident timelines before applying changes to the agent's memory.

Question 2

How does Archivist ensure persistent memory for coding agents?

Accepted Answer

Archivist employs a robust three-tier memory model (working, compact core, and archival graph) with graph and lifecycle safety. It stores data in SQLite, providing optimistic concurrency, immutable audit events, and controlled write workflows to maintain high-quality, persistent memory.

Question 3

What retrieval capabilities does Archivist offer?

Accepted Answer

Archivist features advanced hybrid retrieval, combining full-text search (FTS), local embeddings, graph degree analysis, and recency, all with detailed provenance. This ensures efficient and relevant recall of information for coding agents.

Question 4

Which communication protocols does Archivist support?

Accepted Answer

Archivist supports multiple Model Context Protocol (MCP) transports, including JSON-RPC over STDIO and HTTP. It also offers a secure SSE team mode with authentication for collaborative environments, allowing flexible integration with various MCP clients.

Question 5

What is Archivist and its primary purpose?

Accepted Answer

Archivist is a local-first project memory server designed for coding agents and developers. Its primary purpose is to manage and persist crucial project context—such as decisions, incidents, rules, and graph relationships—in a SQLite database, ensuring continuous context across multiple development sessions.

Archivist

Archivist

主な機能

ユースケース

主な機能

ユースケース