OPEN SOURCE PULSE8.ai Cortex is on the MCP Registry — one vault, every MCP-compatible client
Open-source · Apache 2.0 · MCP-native

The knowledge layer
for AI agents.

Git-native memory, a typed knowledge graph, and MCP-powered retrieval on top of plain Markdown — so agents build, evolve, and reuse persistent knowledge instead of re-deriving it on every query.

Knowledge that compounds.

Ingest· Compile· Link· Retrieve ↺ and it compounds
vault · typed knowledge graph
note note agent session daily feedback
note agent_def session · daily feedback
The problem

Most agents can access tools.
They can't accumulate knowledge.

Traditional RAG retrieves documents and forgets. Cortex builds a persistent, evolving knowledge layer that grows over time — and becomes more valuable the more agents and humans interact with it.

Situation

Knowledge lives in documents

RAG chunks and re-indexes the same sources on every query. Nothing is learned, nothing is kept, nothing connects.

Complication

Every session starts from zero

Insights vanish when the context window closes. Agents re-derive what they already worked out yesterday — and can't build on each other.

Resolution

A knowledge layer that compounds

Every read, write, ingest and compile is logged and graph-linked. Files survive sessions, deployments and model upgrades — diff-friendly Markdown, no database.

What agents can do

A unified vault for agents and humans

Backed by a typed knowledge graph, hybrid search, and a MarkItDown-powered file compiler. Drop files in. Let agents read, write, search, link and compile knowledge.

Learn from interactions

Every read, write, ingest and compile event is logged and graph-linked — the vault learns as it's used.

Build & traverse graphs

Wikilinks, tags and typed edges, maintained automatically. vault_context builds a ranked subgraph from any seed query.

Store structured insights

Markdown notes with typed nodes — note, agent_def, session, daily, feedback.

Retrieve across projects

Hybrid search — BM25 + vector + re-ranking — with keyword and semantic modes, cached with a configurable TTL.

Share through MCP

One vault, every MCP-compatible client — streamable HTTP or stdio, mirrored 1:1 by a REST API at /api/v1/.

Version through Git

The vault is a plain directory of Markdown — diff-friendly out of the box. Zero database; everything persists as Markdown + JSON on your filesystem.

How it works

No database. Just files.

The same vault is reachable over MCP and REST, kept in sync by a real-time filesystem watcher. Everything below the surface is plain Markdown plus a typed graph.

01

Drop files in

PDF, DOCX, PPTX, XLSX, HTML, images and more are converted to Markdown via MarkItDown. Bulk-ingest hundreds at once with SHA-256 dedup.

02

Compile & cross-reference

Raw sources compile into interlinked wiki articles. An LLM is used only to detect cross-references between articles — file conversion needs no API key.

03

Graph stays live

A NetworkX engine maintains wikilinks, tags and typed edges on every file change. The vault watcher keeps the graph in sync automatically.

04

Retrieve & compound

Agents query over MCP; vault_context returns a ranked subgraph. Every operation mirrors into a greppable daily activity log.

MCP
MCP client
Claude Desktop, Cursor, any MCP client
↓  MCP (HTTP or stdio) · REST
:8420
PULSE8.ai Cortex
Auth · MCP /mcp/ · REST /api/v1/ · Vault Watcher · Graph + Compiler
FastAPI
:3100
QMD search
BM25 + vector + re-ranking, runs on-device
hybrid
/vault
The vault
wiki/ · raw/ · agents/ · sessions/ · daily/ · feedback/
Markdown
The interface

Nine MCP tools. One vault.

Every tool is exposed over MCP and mirrored 1:1 in the REST API. Plug into any AI client that speaks the Model Context Protocol.

vault_read

Read a note by path.

vault_write

Create or update a note.

vault_search

Search the vault — keyword, semantic or hybrid.

vault_link

Create, query or delete graph edges.

vault_context

Search → graph traversal → ranked subgraph.

vault_ingest

Ingest raw content or binary files.

vault_compile

Compile raw sources into wiki Markdown.

vault_feedback

Submit feedback on vault quality.

vault_list_feedbacks

List feedback note metadata.

How it compares

Built for knowledge, not just documents

RAG retrieves chunks. GraphRAG adds structure but no memory. Cortex is a persistent, Git-versioned, MCP-native knowledge layer.

Capability PULSE8.ai Cortex Traditional RAG GraphRAG
Persistent knowledge~
Markdown-native storage
MCP-compatible out of the box
Knowledge graph
Git versioning
Agent memory layer~
Human + AI collaboration~
Continuous knowledge evolution~
Zero database required
Editions

Open source to run yourself.
Enterprise when you need scale and governance.

Start free and self-hosted. Bring in managed deployment, single sign-on, compliance guardrails and support when Cortex moves into production.

Open Source

Self-host, forever free

The full knowledge layer, on your own infrastructure. No database, no lock-in.

Free · Apache 2.0

  • Typed knowledge graph (NetworkX) — auto-maintained on every change
  • Hybrid search — BM25 + vector + re-ranking, on-device
  • MCP server + REST API, mirrored 1:1
  • File compiler — PDF, DOCX, PPTX, XLSX, HTML, images
  • Git-native Markdown vault, zero database
  • Bulk ingest with SHA-256 dedup & API-key auth
  • Docker deployment & community support
Enterprise

Governed, supported, at scale

For regulated teams running Cortex in production — part of the PULSE8.AI platform.

Custom · let's talk

Everything in Open Source, plus:

  • Single sign-on with Microsoft Entra ID (OIDC) + MFA
  • Managed or private-cloud deployment
  • GPU-accelerated search at production scale
  • Microsoft Teams notifications & workflows
  • Compliance guardrails & core-system connectors
  • Priority support, SLAs & security review
On the enterprise roadmap Governance & access policies · multi-vault federation · knowledge lineage tracking
Works with

Speaks MCP — plugs into your stack

The vault is just files, so humans and agents collaborate on the same knowledge. No proprietary format, no lock-in.

Claude Desktop Claude Code Cursor VS Code JetBrains IDEs LangGraph LangChain CrewAI AutoGen OpenAI Agents Gemini Obsidian any MCP client any Markdown editor any Git client
Get started

Running in three commands

Cortex requires Docker. An OpenRouter key is optional — needed only for LLM cross-referencing between articles. File conversion works with no key at all.

1

Clone the repository

Grab the source from GitHub and step in.

2

Launch Cortex

start.sh builds Cortex (API + MCP on :8420) and QMD (search on :3100), waits for health checks, and you're ready.

3

Connect your MCP client

Point Claude Desktop — or any MCP client — at http://localhost:8420/mcp/.

bash — pulse8-ai-cortex
# 1 · clone
$ git clone https://github.com/synpulse8-opensource/\
      pulse8-ai-cortex-knowledge-vault.git
$ cd cortex-knowledge-vault

# 2 · launch (Cortex :8420 + QMD :3100)
$ ./scripts/start.sh
 cortex   healthy  → http://localhost:8420
 qmd      healthy  → http://localhost:3100

# 3 · connect your MCP client
 http://localhost:8420/mcp/
Built on great ideas

Standing on open-source shoulders

Andrej Karpathy

The LLM Wiki pattern

A persistent, compounding knowledge base maintained by LLMs — compiled and interlinked incrementally rather than re-derived from raw documents on every query. The direct inspiration for Cortex's architecture.

Tobi Lütke

QMD search

The on-device engine powering all full-text and hybrid search in Cortex — combining BM25, vector search and LLM re-ranking, all running locally.

Microsoft

MarkItDown

The file-to-Markdown converter behind the Cortex compiler — turning PDF, Office docs, HTML and images into structured Markdown for ingestion.

Agents need a knowledge layer —
not another document store.

Star the repo, open a discussion, and help build the knowledge layer for agentic AI.