managed cloud · self-host · source available

The knowledge base your AI tools deserve

A small, human-curated knowledge base for your projects — versioned, searchable, and served to any AI tool over a REST API you control. Claude Code, Cursor, ChatGPT: one source of truth, no vendor lock-in.

Start free → Self-host →

auth-patterns.md v2

auth rails

Architecture Guide v3

architecture conventions models

              # Models

              Use concerns for shared

              behaviors: SoftDeletable,

              Archivable, Pinnable...

Your AI tools forget everything.
Every. Single. Session.

Context engineering has become a core developer skill, yet there are no dedicated tools for it. You're using scattered files and copy-paste as makeshift solutions.

Context Amnesia

You re-explain architecture, conventions, and decisions at the start of every session. Five minutes of context setup, every time.

Context Silos

Knowledge prepared for Claude doesn't transfer to Cursor. You maintain duplicate context across every tool in your stack.

Context Drift

Your architecture decisions evolve, but stale context leads to AI generating patterns you abandoned months ago. No versioning, no history.

Curate once, access everywhere

Organize project knowledge into workspaces. Write versioned markdown memories. Serve them to any AI tool via API.

MEMORY.md for the cheat sheet. recuerd0 for the deep knowledge. Query exactly the context you need.

Organize

Group memories into workspaces — one per project, domain, or team context.

Write

Create markdown memories with tags, versioning, and full-text search.

Serve

REST API with Bearer auth. Point any AI tool at your knowledge, instantly.

Built for developer workflows

Full REST API with token authentication. CLI for terminal access. Agent workflows for persistent AI memory.

Full API Reference → CLI Reference → Agent Workflows → MCP Guide →

rest api

# List workspace memories
GET /workspaces/1/memories.json

# Search across all memories
GET /search.json?q=authentication

# Create a new memory
POST /workspaces/1/memories.json
{
  "memory": {
    "title": "Auth Conventions",
    "content": "# Authentication\n...",
    "tags": ["auth", "rails"]
  }
}

# Version a memory
POST /workspaces/1/memories/5/versions.json
          

cli tool

# Setup your account
$ recuerd0 account add personal --token YOUR_TOKEN

# Search your knowledge
$ recuerd0 search "turbo streams"

# Create a memory
$ recuerd0 memory create --workspace 1 --title "Go patterns" --content "..."

# Read a memory
$ recuerd0 memory show --workspace 1 42

# Pipe content from stdin
$ cat notes.md | recuerd0 memory create --workspace 1 --title "Notes" --content -
          

agent config

# CLAUDE.md agent config
## Memory Server
recuerd0 API: https://recuerd0.ai/api
token: $RECUERDO_TOKEN

## Pre-session
Search workspace "rails-app" for
relevant context before starting.

## Post-session
Save key decisions and patterns
as new versioned memories.
          

mcp

# Add the server in Claude Desktop or Claude.ai
# Settings → Connectors → Add custom connector
https://recuerd0.ai/mcp

# First connect opens a consent screen —
# approve the scopes, no token to copy:
#   Read your memories
#   Create and update memories
#   View your workspaces

# Then ask the model directly:
#   "Search my recuerd0 memories for auth decisions"
          

Everything you need, nothing you don't

Built with Rails 8, SQLite, and a philosophy of simplicity. No external services, no data leaving your network.

Human-Curated

Deliberate, human-in-the-loop knowledge — not automatic capture that accumulates noise. Every memory earns its place: reviewed, versioned, and yours.

human-curated versioned

Version History

Branch from any version with a flat versioning model. Track how your project knowledge evolves — like Git for context.

branching consolidation

Full-Text Search

Keyword search powered by SQLite FTS5 with the trigram tokenizer. Search every workspace in milliseconds, from 3-character queries up. Focused, curated workspaces keep the context you need easy to find.

fts5 trigram

Tool Agnostic

A REST API with Bearer auth works with any tool — Cursor, ChatGPT, your own scripts. For Claude Code there's a native plugin: a skill that captures, searches, and versions session knowledge without leaving the harness. No vendor lock-in, ever.

rest-api cli ai-agent

MCP Server

Connect Claude Desktop, Claude.ai, and any MCP client straight to your memories in the managed cloud. The client registers itself and you approve a consent screen — no tokens to copy, no config files to hand-edit. Six tools cover reading, searching, and writing memories across your workspaces.

remote mcp oauth 2.1 consent-based

Team Sharing

Account-based multi-tenancy with admin and member roles. Invite up to 10 users with signed tokens. Share canonical project knowledge.

roles invitations

Workspace Organization

Pin, archive, soft-delete, and tag. Workspaces are tuned for a curated set of high-value memories — think dozens, curation over accumulation — so the context you serve stays sharp, not noisy. Deep, large workspaces work too; keeping recall sharp as they grow is what semantic search (on the roadmap) is for.

pins tags archive

Linked & Categorized

Type every memory — decision, discovery, preference, or general — and link related ones across workspaces as see-also references. An agent reading one memory can pull the connected context on demand, without re-searching.

categories see-also cross-workspace

Stop re-explaining your project.
Start preserving.

Start in minutes on the managed cloud — or run it on your own server. Your data, your context, accessible from any AI tool.

Start with Cloud → See pricing

or Self-host →