arrow_back Back to AIFC
H
pending Claude

Your Agent Can't Use Context It Can't Find — StashBase Fixes That

Grounded / Real Inflated / Uruttu
70% real
30% uruttu
article Original Content
Your agent can’t use context it can’t find.
StashBase is an open-source, local-first app for turning notes, research libraries, and project docs into a Wiki your agents can use.
It helps you turn messy local folders into reusable agent context by creating linked Markdown Wiki Pages and making sources searchable by meaning as well as keyword.
Key features:
• Linked Wiki Pages – agents can create overviews and focused pages that link back to the original files
• Search across difficult formats – extracts text from PDFs, DOCX files, images, and recordings for meaning and keyword search
• Agent connections – use the included OpenQuill, bring Claude Code or Codex into Chat, or connect an external MCP client
• Files stay under your control – source files keep their layout and Wiki Pages remain ordinary Markdown files you own
• Inspectable workspace – browse, read, and edit supported files beside the conversation, then review tool calls and file edits
It’s open-source (Apache License 2.0 license).
🔗 GitHub: https://lnkd.in/dUwA8wiS
verified Validated Content

Another highly precise one — nearly every bullet in this post matches StashBase's own README almost word for word, though there are a couple of "local-first" nuances worth knowing before repeating that framing uncritically.

The project is real: the canonical repo is liliu-z/stashbase, built by Li Liu (who works on vector retrieval infrastructure at Zilliz/Milvus), Apache 2.0 licensed, with primary support for macOS and Windows and a community-supported Linux build.

Every feature bullet checks out precisely against the project's own docs:

  • Linked Wiki Pages: the project states an agent can create a linked overview and focused pages that explain what's inside, with links back to the original files — matching your bullet almost exactly.
  • Search across formats: confirmed — it extracts searchable text from PDFs, DOCX files, images, audio, and video for meaning and keyword search.
  • Agent connections: confirmed verbatim — use the included OpenQuill, bring Claude Code or Codex into Chat, or connect an external MCP client to the same library.
  • Files stay yours: confirmed — sources keep their original layout; Wiki Pages are visible Markdown you can open and use outside StashBase.
  • Inspectable workspace: confirmed — tool calls and file edits can be reviewed in the app.
  • License: confirmed Apache 2.0.

One nuance worth flagging on "local-first": keyword search and file browsing are genuinely local and need no cloud account, but semantic ("search by meaning") search is not fully local — it sends relevant text to a selected embedding provider for indexing and queries, either via StashBase's hosted service or your own OpenAI/OpenRouter key. Similarly, OpenQuill runs locally, with prompts and necessary model context sent through StashBase's hosted model gateway. None of this contradicts "open-source" or "files stay under your control" (the files themselves do stay local), but "local-first" could read as more fully offline than the semantic-search feature actually is.

One other thing worth knowing: official desktop builds enable basic usage-statistics collection by default (opt-out available in settings) — a minor point, but relevant context for a tool marketed on local/user-controlled data.