Wenlan vs Basic Memory: Source-Backed AI Work vs Shared Markdown Knowledge

Choose Basic Memory if you want people and AI assistants to work in the same Markdown knowledge base. You can keep it local, add Basic Memory Cloud for hosted access and sync, or use Team workspaces for shared knowledge.

Choose Wenlan if the harder problem is turning AI work into maintained knowledge: capture decisions across sessions, hand work off between agents, and distill supported context into source-backed Pages that remain reviewable.

Both expose context through MCP, keep important material human-readable, and support semantic search. The meaningful difference is the operating model around that context.

Basic Memory's open-source path stores human-readable Markdown files and builds a secondary index for search and graph operations. Its MCP server lets supported AI clients read and write that knowledge base, while the files remain usable in editors such as Obsidian or VS Code.

Basic Memory Cloud is the hosted path. It adds remote MCP access, a web editor, optional local sync, snapshots, and file history. Team workspaces add shared projects, membership, collaborative editing, activity, and per-file history.

Current Basic Memory also provides semantic search, graph traversal, `build_context`, and Agent Skills that teach assistants to search before answering, capture durable knowledge, link related notes, and maintain the knowledge base. It is no longer accurate to describe the product as only a local vault or as having no work-loop guidance.

Wenlan starts from the work an AI agent is doing. Agents explicitly capture decisions, lessons, and corrections as Memories; `/handoff` preserves the state of an unfinished session; `/distill` turns repeated, supported context into Pages; `/curate` and `/lint` make review and repair visible.

Sources are the evidence layer. Memories retain durable working context. Pages synthesize that context into readable knowledge with source IDs, citations, and revision history. The distinction is deliberate: recalled fragments and maintained wiki pages do not silently become the same thing.

The local daemon provides retrieval across MCP clients. Readable Pages, sessions, and status artifacts are projected under `~/.wenlan/` with local git history, while capture and retrieval remain backed by the daemon-owned store.

Imagine that three coding agents investigate the same authentication problem over a week. With Basic Memory, the durable object is the shared knowledge base: agents search existing notes, add observations, link related entities, and update a Markdown document that the team can also edit.

With Wenlan, the durable path is evidence to maintained knowledge: each agent captures the decisions or gotchas it learned, hands off unfinished state, and later distills supported memories into a Page whose source chain can be inspected and revised.

Basic Memory is the more direct fit when the shared note is the product. Wenlan is the more direct fit when you need to see how session evidence became a maintained answer.

For local Basic Memory, Markdown files are the primary human-readable record and a database acts as a derived search index. File history depends on the local tools you choose. Basic Memory Cloud adds managed sync, snapshots, and hosted per-file history; Teams extends that hosted model to collaborators.

Wenlan keeps raw captures in its local daemon store for recall and projects readable Pages, sessions, and status Markdown under `~/.wenlan/`. Those projected artifacts are versioned locally in `~/.wenlan/.git/`, so Page revisions can be inspected without turning the projection into the retrieval database.

Neither shape is automatically better. Basic Memory favors a file-first knowledge base that can gain hosted collaboration. Wenlan favors a local retrieval store plus human-readable, source-linked outputs.

Basic Memory combines text and semantic search with graph traversal over observations and relations. Its `build_context` tool assembles connected knowledge, and its Agent Skills give assistants an explicit search, capture, and maintenance routine.

Wenlan combines full-text and embedding retrieval, weighted fusion, eligible graph context, and optional reranking. Its plugin workflow separates `/recall`, `/capture`, `/handoff`, `/distill`, `/curate`, and `/lint`, so retrieval, capture, synthesis, review, and repair remain distinct actions.

The products publish different retrieval evidence, so this page does not turn unmatched benchmarks into a winner. Test each system with the same real task: recover an old decision, update it after a correction, and inspect why the final answer should be trusted.

Basic Memory makes notes and their relationships inspectable. A team can read the Markdown, follow links, inspect Cloud history, and decide what belongs in its shared knowledge base.

Wenlan adds a stricter boundary between working memory and a distilled Page. A Page record carries source memory IDs, citations, provenance state, and revisions, and the daemon rejects a Page with no source. That is useful when the question is not only “what does the note say?” but “which captured evidence supports this maintained claim?”

Basic Memory can still hold source citations, and Wenlan Pages remain editable and reviewable. The difference is which provenance behavior the system requires rather than what a careful author could add manually.

Choose Basic Memory when you already think in Markdown notes, want AI assistants to share that knowledge base, or need a hosted web and Team path without building your own sync layer.

It is also the clearer fit when people and agents should directly co-edit the same durable notes and the knowledge base—not session provenance—is the primary unit of work.

Choose Wenlan when context is scattered across AI sessions and tools, and you want explicit capture and handoff before that context is promoted into maintained knowledge.

It is the clearer fit when source-backed distillation, reviewable revisions, local operation, and a visible repair loop matter more than hosted team collaboration.

Moving from Basic Memory to Wenlan is selective today. Choose the durable notes that should become Sources or Memories, capture them through the CLI or MCP workflow, and distill only the claims whose source chain you want Wenlan to maintain. Wenlan does not currently advertise a one-command Basic Memory vault importer.

Moving Wenlan's readable output into another knowledge base is straightforward at the file level: Pages and sessions under `~/.wenlan/` are Markdown. That preserves readable content, but not Wenlan's live recall, provenance state, curation, or distillation behavior.

Using both is technically possible because each exposes an MCP server. If you do, assign ownership clearly—for example, Basic Memory for shared team notes and Wenlan for source-backed personal AI-work history—so agents do not create conflicting copies.