Skip to content

How Neither works

Scattered notes.
Connected context.

Neither turns scattered project notes into a structured context graph, ready for your AI through MCP.

  1. 01 / Your sources

    Add the pieces.

    Save a note, import documents, or connect an available work tool.

  2. 02 / Neither organizes

    Build project context.

    Extracted facts merge into existing records and a shared project hierarchy.

  3. 03 / MCP retrieves

    Bring the connections.

    Your AI receives relevant records, supporting excerpts, and source references.

  4. 04 / Your AI answers

    Use the evidence.

    Your AI combines the retrieved evidence into an answer you can check.

What the graph adds

The build is ready.
Is the launch?

The answer depends on more than the build note. Neither keeps the project’s facts together, with the material behind them.

Simplified project record

Beacon launch

Build · Launch plan
Build ready.
Requirement · Requirements
Security approval required before launch.
Review · Security review
Approval pending. Maya owns the security review.

Your question

“Can we launch?”

Your AI’s answer

Not yet.

Security approval is missing. Maya owns the review.

The AI combines the launch requirement and review status returned by Neither. The returned source references let you check the conclusion.

Explore the connected example →

Fictional example. Extraction and answers can be incomplete; check the sources when the details matter.

You decide what goes in

Start with what you have.

Save a note

Ask your AI to add it.

Explicitly ask your connected AI to save context using Neither. Only the content sent by that tool call is added; the rest of the chat is not automatically recorded.

Import documents

Bring existing notes.

Push selected Markdown and text files with the CLI. Push again after edits, or use the folder watcher while it is running.

Document and CLI guide →
Connect a source

Keep project context current.

Available work-tool connections import supported information and updates. Coverage, initial history, and sync timing depend on the provider.

See source coverage →

Use it through MCP

Inside your AI tool.

MCP is the connection that lets your AI add context to Neither and retrieve structured records with their sources.

Cursor and Claude Desktop connect with a local MCP server (stdio, Node.js 20 or later). ChatGPT connects with the hosted MCP URL https://api.neither.online/mcp in developer mode on the web. ChatGPT Pro can retrieve Neither context; saving from ChatGPT requires Business, Enterprise, or Edu developer mode. Some other HTTP-only hosts may still be unable to connect.

Open guided setup →

Processing comes before retrieval.

New context is processed before it becomes searchable. Setup shows progress and verifies that your AI received a result with a source.

What you can retrieve has limits.

Free includes 10 MB of context, 1000 reads a day, and 3-day connector history. Missing sources and plan limits affect the context available.

See current plans →

Ready when you are

Give your next conversation
a place to start.

Begin with one project note. See what your AI can recover.

We use privacy-friendly analytics to improve the public site. No tracking runs until you accept. Essential operational telemetry stays on either way.