The company brain for AI agents
Neither turns email, chat, docs, meetings, CRM, and tickets into a living, cited map of people, decisions, owners, sequence, and dependencies — then gives each agent only the context it needs.
RAG finds matching text. Neither maintains what happened, who owns it, and what it blocks.
Free beta — no waitlist. 10 MB context, 500 memory reads a day.One living picture of the work
Maya Chen owns the security review blocking Milestone 3.
Who owns it. What it blocks. What changed.
Why this is not another RAG layer
RAG is useful for document retrieval. Long-running agents need maintained operating state: owners, sequence, decisions, blockers, and evidence.
Who owns the security review, and what's it blocking?
Several matching chunks enter the prompt. The model must infer the owner, timeline, and dependency again.
Maya Chen (Security Lead, Trident Corp) — last mentioned Feb 8, no reply to Ahmed's follow-up. Blocks Milestone 3 of the Trident contract, 5 days overdue.
Context-window advantage
Retrieval can keep adding loosely related chunks to the prompt. Neither maintains the relationships as work arrives and returns one bounded context pack — enough to reason and act without dragging the document pile into every turn.
Decision recall: 98% vs 44% for naive vector search, across 50 held questions.
Published comparison uses a documented naive vector-search baseline on a pinned corpus — not a claim that every RAG architecture behaves the same. See /benchmarks for the method.
One maintained brain
MCP, the Context API, and Neither itself all receive context from the same maintained graph.
Email, chat, docs, meetings, CRM, and tickets feed one company brain.
People, decisions, ownership, sequence, dependencies, and evidence stay linked.
Each request gets a bounded context pack with citations, freshness, and workspace scope.
We use privacy-friendly analytics to improve the public site. No tracking runs until you accept. Essential operational telemetry stays on either way.