Via — by Vektor Memory

Route anything. Remember everything. Works everywhere.

https://medium.com/@vektormemory/all-roads-lead-to-ai-rome-1c601f1ec440

Via is the universal integration layer for AI tools. One CLI that connects Claude, Cursor, Windsurf, and ChatGPT to a shared memory, task board, and context bus — so your work follows you across every tool, every session, every machine.

Part of the Vektor ecosystem alongside Vex and Slipstream.

The problem

Every AI tool remembers inside its own walls. Claude forgets what you did in Cursor. Cursor forgets what you built in Windsurf. The moment you switch tools — or open a new session — context resets to zero.

Via is the bus between them.

Install

npm install -g @vektormemory/via
via --help

Requirements: Node.js >= 18. No native dependencies.

What makes Via different

`via diff` — Compare AI tools side by side

The feature no other tool has. Ask the same question to Claude and Cursor, then see exactly where they agree, diverge, and what unique concepts each one brought.

via diff "explain microservices"
via diff add claude "Microservices split apps into small independent services..."
via diff add cursor "Microservices are small focused services that communicate via APIs..."
via diff show

┌─ DIFF — explain microservices ────────────────
│ claude          12 words
│ cursor          14 words
│ similarity      21% word overlap
│
│  claude                          |  cursor
│  ──────────────────────────────  |  ──────────────────────────────
│  Microservices split apps into   |  Microservices are small focused
│  small independent services...   |  services that communicate via...
│
│ claude unique terms  independent, database
│ cursor unique terms  focused, communicate, deployed
└───────────────────────────────────────────────

`via memory` — Relationship-aware code indexing

Point at any folder and Via extracts symbols, imports, and function definitions from 10+ languages — then builds an import graph in SQLite. Search traverses relationships, not just text.

via memory add --file ./src/
# → extracts symbols + import edges from JS/TS/Python/Go/Rust
# → via memory search "auth" returns auth.js + every file that imports it

┌─ MEMORY — SEARCH: auth ───────────────────────
│
│  Direct matches (2 files)
│    ● auth.js       ./src/auth.js
│    ● config.js     ./src/config.js
│
│  Related via imports (3 files)
│    ○ server.js     ./src/server.js
│    ○ middleware.js ./src/middleware.js
│    ○ routes.js     ./src/routes/auth.js
└───────────────────────────────────────────────

No embeddings. No API calls. Pure local SQLite graph.

`via convert` — Local file conversion pipeline

Convert any file locally — audio, video, images, documents, archives — powered by FFmpeg, ImageMagick, Pandoc, and LibreOffice. Nothing uploaded anywhere. Pipe output directly into via memory with --ingest.

via convert ./report.pdf --to md                  # pdf → markdown
via convert ./recording.mp3 --to txt              # audio → transcript
via convert ./video.mp4 --to gif                  # video → gif
via convert ./doc.docx --to pdf                   # office → pdf
via convert ./report.pdf --to md --ingest         # convert + store in memory
via convert --check                               # check installed tools
via convert --formats                             # all supported formats

┌─ CONVERT — TOOL CHECK ────────────────────────
│ FFmpeg          installed  audio/video
│ ImageMagick     installed  images
│ Pandoc          installed  documents
│ LibreOffice     installed  office docs
│ Poppler         installed  pdf→txt
│ Zip             installed  archives
│ 7-Zip           installed  7z archives
└───────────────────────────────────────────────

Commands

`via init`

Wire via into every AI tool detected — one command, fully configured.

via init           # writes MCP config for Claude Desktop, Cursor, Windsurf
via init --dry-run # preview what would be written

`via memory`

Fact storage + relationship-aware codebase indexing.

via memory add "JWT tokens expire in 1h"          # store a fact
via memory add --file ./src/                      # index a codebase
via memory search "auth"                          # search + related files
via memory graph                                  # show import relationships
via memory list                                   # list indexed files + facts

`via convert`

Local file conversion. No uploads. Optional memory pipeline.

via convert <file> --to <format>                  # convert a file
via convert <file> --to md --ingest               # convert + store in memory
via convert --check                               # check installed tools
via convert --formats                             # show all supported formats

Supported formats:

Images     png jpg webp gif bmp tiff ico svg  →  png jpg webp gif bmp tiff ico pdf
Audio      mp3 wav ogg m4a aac flac aiff wma  →  mp3 wav ogg m4a aac flac aiff
Video      mp4 mkv mov avi webm flv wmv       →  mp4 mkv mov avi webm gif mp3
Documents  md rst html txt tex org epub       →  md html txt pdf epub docx odt
Office     docx doc odt rtf xlsx pptx         →  pdf txt html odt docx
PDF        pdf                                →  txt md html docx
Archives   any file or folder                 →  zip tar.gz 7z

External tools required: FFmpeg, ImageMagick, Pandoc, LibreOffice, Poppler, 7-Zip.

`via task`

Shared persistent task board any AI tool can read and write via MCP.

via task add "refactor auth module" --high
via task
via task done <id>
via task start <id>

`via log`

Unified activity log. Auto-captures Claude Code sessions.

via log "decided to use postgres" --tool claude
via log --scan     # one-shot capture of all Claude Code sessions
via log --watch    # live capture as sessions complete
via log --today
via log search "postgres"

`via ask`

Route a question to the right AI tool — and open it.

via ask "should I use postgres or sqlite?"        # opens recommended tool
via ask "refactor auth module" --tool cursor      # force a specific tool
via ask "explain this architecture" --no-open     # recommend only

`via diff`

Compare what two AI tools said about the same prompt.

via diff "your prompt"          # register a new prompt
via diff add claude "..."       # store Claude's response
via diff add cursor "..."       # store Cursor's response
via diff show                   # side-by-side + unique terms
via diff list                   # all saved comparisons

`via handoff`

Export your full working state before switching tools.

via handoff --export                        # saves .vstate.json
via handoff --import ./sprint3.vstate.json  # restore on any machine
via handoff --list

`via serve`

Run Via as an MCP server. Claude Desktop, Cursor, and Windsurf can call via tools natively.

via serve           # stdio (Claude Desktop, Cursor)
via serve --sse     # HTTP+SSE mode

Claude Desktop config:

{
  "mcpServers": {
    "via": { "command": "via", "args": ["serve"] }
  }
}

Or just run via init — it writes this automatically.

MCP Tools (via serve)

When running as an MCP server, Claude Desktop and Cursor can call:

via_task_list      List open tasks
via_task_add       Add a task
via_task_done      Mark task done
via_memory_add     Store a fact
via_memory_search  Search stored facts (relationship-aware)
via_log            Log a decision or event
via_context        Pull formatted memory context
via_status         Ecosystem health check

Vektor ecosystem

Via          Route context and execution across all AI tools
Vex          Migrate agent memory between vector stores
Slipstream   Graph memory, vector search, multimodal

Via uses SQLite locally. No embeddings, no API calls for indexing. When you need semantic search, graph traversal, or team-shared context — upgrade to Vektor Slipstream.

Roadmap

v0.3 — current

via init, via memory, via convert, via task, via log, via ask, via diff
Relationship-aware codebase indexing — symbols + import graph in SQLite
Local file conversion pipeline — FFmpeg, Pandoc, LibreOffice, ImageMagick
Claude Code session auto-capture
MCP server with 8 tools
Claude Desktop, Cursor, Windsurf support

v0.4 — coming

via memory semantic search upgrade via Slipstream
Team-shared task board
via diff --live real-time comparison
via convert batch mode — convert entire folders

License

Apache 2.0 — free forever. Built by Vektor Memory.

Name		Name	Last commit message	Last commit date
Latest commit History 16 Commits
commands		commands
connectors		connectors
utils		utils
CHANGELOG.md		CHANGELOG.md
LICENSE		LICENSE
README.html		README.html
README.md		README.md
SPEC.md		SPEC.md
package-lock.json		package-lock.json
package.json		package.json
via.mjs		via.mjs

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Via — by Vektor Memory

The problem

Install

What makes Via different

`via diff` — Compare AI tools side by side

`via memory` — Relationship-aware code indexing

`via convert` — Local file conversion pipeline

Commands

`via init`

`via memory`

`via convert`

`via task`

`via log`

`via ask`

`via diff`

`via handoff`

`via serve`

MCP Tools (via serve)

Vektor ecosystem

Roadmap

License

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Folders and files

Latest commit

History

Repository files navigation

Via — by Vektor Memory

The problem

Install

What makes Via different

via diff — Compare AI tools side by side

via memory — Relationship-aware code indexing

via convert — Local file conversion pipeline

Commands

via init

via memory

via convert

via task

via log

via ask

via diff

via handoff

via serve

MCP Tools (via serve)

Vektor ecosystem

Roadmap

License

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

`via diff` — Compare AI tools side by side

`via memory` — Relationship-aware code indexing

`via convert` — Local file conversion pipeline

`via init`

`via memory`

`via convert`

`via task`

`via log`

`via ask`

`via diff`

`via handoff`

`via serve`

Packages