For AI Agents

Graph + temporal memory for MCP-capable agents — Claude Code, Cursor, n8n, Continue, Goose. Local-first or EU-hosted.

Join early access

Want the notes app instead? See the human side of Korely →

korely_memory.py python 
# 1. Connect
from korely_memory import Korely
korely = Korely(api_key="kor_live_...", region="eu")

# 2. Remember (graph + temporal built-in)
korely.add(
    "User prefers TypeScript with strict mode",
    agent_id="coding-assistant",
)

# 3. Recall — handles contradictions automatically
korely.add("Actually I switched to Rust")
result = korely.search("user's preferred language")
# → "Rust" (TypeScript marked as superseded)

Python today, TypeScript and Go in early access. Get notified →

How it works

Three steps, then every agent sees your vault

01 Install

Get Korely or grab the CLI

The Mac app ships the desktop vault and the MCP server stdio binary in the bundle. The standalone CLI installs from npm for headless servers, CI runners, or anywhere you do not want a GUI. Join the early access waitlist below to get the build.

02 Connect

Paste one config into your client

For Claude Desktop, edit claude_desktop_config.json. For Cursor, paste into ~/.cursor/mcp.json. Same shape, same five read tools either way.

03 Share memory

Every agent gets its own folder

Notes saved by Claude land in Agents/Claude Code/. Cursor in Agents/Cursor/. n8n in Agents/n8n/. Same vault, traceable provenance.

MCP clients

Plug Korely into the tools you already use

Claude logo

Claude Code

Anthropic

Persistent project memory across coding sessions.

See use cases
Cursor logo

Cursor

Anysphere

Memory between PRs, decisions saved as notes.

See use cases

n8n

n8n

Workflow runs write their context, next runs read it back.

See use cases
Continue.dev logo

Continue

Continue.dev

IDE assistant with shared codebase memory across tabs.

See use cases
Goose logo

Goose

Block

Persistent task history for Block's agent.

See use cases

Any MCP-aware client.
The protocol is open.

CLI

Same vault from a shell, fraction of the tokens

MCP is great when you want an agent to explore. The CLI is what you reach for when you already know what to pull. Same five read tools, invoked from any terminal, pipeable into anything.

Useful for cron jobs, CI prep, editor commands, shell scripts, and any workflow where you want to drive the query yourself instead of waiting for an agent to. More on the CLI → 

# Find every note about a topic, top 5
korely search query="Q3 pricing decisions" limit=5

# Pipe the result to a local LLM for free
korely search query="customer feedback" \
  | ollama run llama3.1 "summarize"

# Cron: dump open actions every Monday as JSON
korely search query="open action items" \
  --format=json > weekly.json

# Run against a remote Pro cloud vault
korely search query="meeting recap" --remote

Your vault

Your agents' memory is a folder you can cd into

Plain Markdown on your disk. One folder per agent. Open it in Finder, in your editor, or pipe it through grep.

~/Documents/Korely finder 
~/Documents/Korely/
└── Agents/
    ├── Claude Code/
    │   ├── code-conventions.md
    │   └── architectural-decisions.md
    ├── Cursor/
    │   ├── pr-review-notes.md
    │   └── refactor-history.md
    ├── n8n/
    │   └── workflow-customer-data.md
    └── Personal/
        ├── meeting-2026-05-14.md
        └── q3-priorities.md

Grep it.

Every memory is a Markdown file. Standard Unix tools work. No SDK required to read.

Edit it.

Open the vault in Obsidian, Cursor, or any Markdown editor. Korely re-indexes on save.

Move it.

Copy the folder to a USB drive. Sync it with iCloud. The vault travels with you.

What Korely does

One memory layer, every shape your agent needs

Vector, graph, and temporal memory in one place. Markdown on disk. MCP and CLI for any agent.

Capability
How Korely does it
Memory model
Vector, graph, and temporal memory in one store. Contradictions resolved on the fly.
Where memories live
Plain Markdown files on your disk, under a folder you choose.
Agent integration
Native MCP server (stdio today, cloud in early access). Plus a CLI for shell-driven workflows.
Search
Hybrid: keyword (FTS) and semantic (local embeddings), fused with Reciprocal Rank Fusion.
SDKs
Python at launch. TypeScript and Go in the early-access cohort.
Hosting
Local-first by default. Optional EU servers (Helsinki) when you want to share a vault across machines.
Pricing model
Flat indie pricing. Graph and temporal included in every paid tier.

Use cases

Six shapes of agent memory

One vault, scoped per agent. Each verticale ships with a code snippet you can copy.

01 For Cursor / Claude Code / Continue

Coding agent

Code conventions and refactor history that survive a new chat. The agent learns how your team writes once, not every Monday.

See the snippet

02 For SDR / AE tools

Sales co-pilot

Deal stages, contact preferences, objections from the last 3 calls. Context that compounds across weeks of a complex pipeline.

See the snippet

03 For helpdesk / chat tools

Support agent

Customer history, past tickets, preferred resolution channel. The agent recognises returning customers and skips the warm-up.

See the snippet

04 For EdTech apps

Tutor agent

Student progress, recurring mistakes, learning style. The agent adapts to each learner without re-discovering them every session.

See the snippet

05 Regulated, EU-hosted

Healthcare assistant

Patient context with audit logs, EU residency, lawful-basis tags. Local-first or on-prem deployment for regulated environments.

See the snippet

06 For consumer companions

Personal AI

Long-term context across years. Color preferences that shift, jobs that change, goals that evolve. Korely tracks the timeline.

See the snippet

Pricing

Flat indie pricing, with the graph included

Free for hobby projects. $29/mo for serious builders. EU-hosted by default.

Hobby

Single dev, side projects.

Free forever
  • 10K memories
  • 1 agent
  • Graph + temporal
  • MCP support
Most popular

Developer

Graph + temporal included.

$29 /month
  • 100K memories
  • 10 agents
  • Graph + temporal
  • MCP support

Team

For growing teams.

$99 /month
  • 1M memories
  • 100 agents
  • Graph + temporal
  • MCP support

Paid tiers launch in Q3 2026. Join early access to get the first year at half price.

Early access

Stop renting your agent's brain

Python SDK ships in Q3 2026. Leave your email and you'll be in the first cohort, with the first paid year at half price.

One email when the SDK ships. No newsletter, no tracking. Unsubscribe anytime.