Protocontext

ProtoContext

Deterministic knowledge layer for RAG systems.

Open standard + search engine + semantic retrieval for AI-readable web content.

ProtoContext lets AI agents read structured knowledge from context.txt instead of depending only on raw HTML scraping and fully probabilistic retrieval pipelines. You can publish context.txt on your domain, or upload structured context directly from the dashboard (no URL required).

Use ProtoContext as a fast deterministic path, and add semantic/vector retrieval only when needed.

Website: https://protocontext.org/

[](SPEC.md) [](engine/) [](LICENSE) [](mcp-server/) [](https://www.npmjs.com/package/n8n-nodes-protocontext) [](https://www.npmjs.com/package/n8n-nodes-protocontext)

[Specification](SPEC.md) • [Quick Start](#quick-start) • [Search Engine](#search-engine) • [n8n Node](#n8n-node)

---

Why ProtoContext

ProtoContext is best when knowledge is structured and deterministic:

• products, prices, hours, policies, docs
• low-latency agent responses
• predictable answers from publisher-authored content
• reduced operational overhead (scraping/chunking/embedding pipeline becomes optional)

Traditional RAG is still best for:

• massive, unstructured corpora
• deep semantic discovery across noisy text
• use cases where exact schema is unavailable

ProtoContext does not replace RAG. It makes full RAG pipelines optional for many production flows.

---

Fast Path vs Traditional RAG

| | Traditional RAG Pipeline | ProtoContext Fast Path | |---|---|---| | Primary input | Unstructured pages/docs | Structured context.txt | | Typical latency | ~200-500ms | One-line description of what this site does

@lang: en @version: 1.0 @updated: 2026-02-25 @topics: your, relevant, topics @content_type: website

section: About

What your site does, in plain text.

section: Key Information

Prices, hours, contacts, features.

Read the full spec: [SPEC.md](SPEC.md)

---

## Search Engine

ProtoContext includes a search engine for AI agents:
- indexes `context.txt` (or falls back to scraping when missing)
- serves structured results in the deterministic fast path
- supports semantic retrieval when configured

### Content Ingestion Priority

| Priority | Source | Method | AI Key Needed |
|---|---|---|---|
| 1 | `/context.txt` | Direct parse | No |
| 2 | `/llms-full.txt` or `/llms.txt` | AI conversion | Yes |
| 3 | `sitemap.xml` | Scrape + convert | Yes |
| 4 | Internal links | BFS crawl fallback | Yes |

Scraping is optional fallback when structured context is not available.

---

## API

All protected endpoints require `X-Proto-Token`.

| Method | Endpoint | Description |
|---|---|---|
| `GET` | `/search?q=...` | Search across indexed sites |
| `GET` | `/site?domain=...` | Get sections for one domain |
| `POST` | `/submit` | Submit domain for indexing |
| `POST` | `/delete` | Remove domain |
| `POST` | `/batch` | Multi-query search |
| `GET` | `/stats` | Index stats |
| `GET` | `/health` | Health check |

Common filters: `domain`, `lang`, `content_type`, `section`, `limit`.

---

## n8n Node

Community node for [n8n](https://n8n.io) — use ProtoContext as a workflow node or as an AI Agent tool.

### Install

In n8n: **Settings → Community Nodes → Install** → `n8n-nodes-protocontext`

Or manually:

```bash
cd ~/.n8n/nodes && npm install n8n-nodes-protocontext

Operations

| Operation | Description | |---|---| | Search | Search across all indexed content | | Get Site | Get all sections for a domain | | Submit Domain | Submit a domain for indexing | | Upload Content | Upload raw context.txt | | Delete Domain | Remove a domain from the index | | Stats | Index statistics |

AI Agent Tool

Connect the ProtoContext node to the tools input of an AI Agent node. The agent will use it to search and retrieve structured knowledge automatically.

📦 npm package

---

Architecture

protocontext/
├── engine/             # API + indexer + crawler/scraper + auth
├── web/                # Dashboard (search, submit, setup, settings)
├── mcp-server/         # MCP tools for AI agents
├── n8n-node/           # n8n community node (workflow + AI Agent tool)
├── validator/          # context.txt validation
├── wordpress-plugin/   # WordPress + WooCommerce integration
├── examples/           # Example context files
└── SPEC.md             # Spec v1.0

Spec is stable (v1.0). Engine is in beta (v0.1.1-beta).

---

License

Apache 2.0. See [LICENSE](LICENSE).

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: protocontext
• Source: protocontext/protocontext
• License: Apache-2.0
• Homepage: https://protocontext.org/

Install and usage instructions live in the source repository linked above.