Start here
Build on your knowledge.
Bring your documents, find the relevant passages, and ask questions with sources you can inspect.
What is ContextPier?
ContextPier connects document ingestion, search, and grounded answers in one workspace. Use the dashboard for everyday work, or the Go HTTP API and authenticated CLI for integration.
An answer is useful when you can check where it came from. Source references connect an answer to immutable document versions and their original passages.
A small, explicit workflow
Documents pass through indexing before they become searchable. Questions use the published sources you currently have permission to read.
- Add a text, Markdown, or PDF document.
- Follow its parse and indexing jobs until publication succeeds.
- Search published passages or ask a question.
- Open the cited passages to inspect the evidence.
Choose your interface
The dashboard provides Documents, Jobs, Search, and Conversations screens. Conversations are saved independent questions and answers.
The Go API provides the same underlying document, retrieval, source, and question operations. The CLI supports authenticated use without a browser.
A context layer for agents
The target architecture connects AI agents to permission-aware, sourced context through a shared MCP gateway. A context packet returns through the same gateway; the caller decides how to use it.
Today, document ingestion, access checks, hybrid search, original-span context and grounded answers run through the Go API, CLI and dashboard. The MCP gateway, external source connectors, dedicated reranking and semantic compression are planned additions.
Evidence comes first
An unrelated question can return insufficient_evidence. This means the available published sources do not provide enough support for an answer.
Publication, retrieval relevance, and answer quality are different checks. A successful upload alone does not prove a document is searchable or that a question can be answered.