The Cursor for writing

Cleve is a beautiful, markdown-based second brain that your whole team can collaborate on live. Use Cleve to give context to Claude, Codex, Cursor, OpenClaw and your own apps.

Start for free Book a demo

Cleve

brainstorm/dog-vacations.md

Cleve

OpenClaw

AI for dog vacationsLizzie

Edited just now

Brainstorming what this could be. Drop ideas in either bucket — the wilder, the better. Nothing's off the table yet.

Ideas

Match each dog's personality to a destination vibeAshvin
Auto-book pet-friendly hotels with verified yardsClaude
Daily photo updates from the doggy hotel

Things to figure out

Liability when the AI books the wrong size crateRashad
Subscription, per-trip, or take a cut of the booking?
Do owners actually want this, or just a human concierge?

Works with

Claude Code

Cursor

Codex

Claude Desktop

OpenClaw

›

Quick start

HTTP · MCPβ BETA

# One install. OAuth on first connect. Same workspace your team already uses.

$claude mcp add cleve --transport http https://app.cleve.ai/mcp

HTTP transport, OAuth via Clerk on first connect. Once installed, your agent has the same access to the workspace your team does — the same notes, folders, skills and memory. Switch clients freely; re-running the command is idempotent.

From the agent's seat

Your agent reads the same file your team just wrote.

When Sam finishes a paragraph in Cleve, your coding agent's next call returns the version Sam just left. No batch sync, no copy-pasted context, no stale snapshot — same source, same instant, both sides of the loop.

Claude Code

MCP · cleve

youImplement the pricing migration.

Reading the launch plan first.

cleve_search_notes("launch plan")

→ 1 match · launch-plan.md

cleve_read_note("launch-plan.md")

# Q1 launch plan

## Week 1 — soft rollout

- Internal beta with the ops team

- Pricing copy approved by Sam

- Suggested by Maya's agent: ship to /pricing-v2 first, then alias

- Comms ready for partner orgs

agentGot it. Maya's agent suggested /pricing-v2 first, then alias the old route. Setting that up.

Try saying

Real prompts. Real tool calls.

Whatever you'd ask a teammate, ask your agent. The MCP layer turns it into the right calls — no glue code, no scraping.

“Push everything in ~/notes/ into Cleve under 'Personal'.”

create_foldercreate_note

“Sync my Obsidian vault with Cleve so the team can edit.”

list_folderscreate_noteedit_note

“Pull the latest product spec into this repo.”

search_notesread_note

“Grab the brand guide and quote the voice rules in this landing page.”

search_notesread_notecite

Beyond markdown

Same MCP. Four surfaces your agent can reach for.

The Cleve MCP server isn't just CRUD on notes. The marketplace, memory and org primitives are exposed too — same shape your team already uses.

Notes & folders

Read, search, write and organise the workspace. Cite specific passages back. The bread-and-butter surface.

search_notesread_notecreate_noteedit_notemove_noteshare_notelist_folderscreate_foldercite

Skills marketplace

Browse, install, write and publish AI skills. Your agent inherits the same library your team curates.

search_marketplaceinstall_skillcreate_skillpublish_skilllist_skills

Memory

Durable workspace memory the agent can read, write, toggle or wipe. Stop reseeding context every conversation.

get_memorytoggle_memoryclear_memory

Organizations

Multi-tenant from day one. List the orgs a user belongs to and switch between them mid-session.

list_orgsswitch_org

Connect your agent.

Run the install command for your client, sign in once, and your agent has the same access to the workspace your team does.

Start for free Book a demo