atmina
Get access →
MCP-native shared memory →

Memory, Shared.

Atmina gives your AI agents a shared cloud memory. Create a knowledge base once, connect it to MCP-compatible harnesses, and let context move with you across sessions and collaborators.

Get access → See how it works

● No credit card · Works with any MCP-compatible agent

~ / connect-atmina 
# Add Atmina to an MCP-compatible client:
{
  "mcpServers": {
    "atmina": { "url": "https://mcp.atmina.ai/mcp" }
  }
}

→ shared memory available · read · write · search
MCP native
Connects as an agent tool
Shared KBs
Memory you can grant access to
Cloud edge
Hosted close to the harness
Audit trail
Tool calls recorded for review
Works with MCP-compatible harnesses
What's inside

One shared memory your agents can actually use.

Atmina combines cloud knowledge bases, MCP tools, hybrid search, and sharing controls so people and agents can work from the same memory.

01 — Shared Knowledge Bases

A cloud memory you can share with people and agents.

Create a KB for a project, a family plan, a client, or your own working notes. Invite collaborators when the memory should be shared, or keep it private when it should stay personal.

  • Private, shared, and public knowledge base visibility
  • Role-based access for viewers, editors, and admins
  • Plain-text read, write, append, and patch tools over MCP
shared memory private · shared · public
atmina shared memory
02 — Context-Aware Retrieval

The right context, the moment your agent needs it.

Atmina searches the files your agents can access and returns ranked context back through MCP. Hybrid retrieval helps agents find the right source material without turning memory into another chat window.

  • Keyword, semantic, and hybrid search modes
  • Results include source metadata for follow-up reads
  • Public KBs support anonymous keyword search
context delivery ranked · sourced
atmina · session resumed + 3 memories
What did we decide about the launch plan?
I found the launch notes and access plan. Signup capture lives on the site, and MCP OAuth is gated by email. ↳ 3 results 2 files read_file next
context: semantic keyword hybrid ranked
03 — Harness Portability

One memory, every harness you bring to it.

Use Claude for one task, Codex for another, then hand the same KB to the next MCP-compatible client. The memory is not trapped inside one chat window.

  • One MCP endpoint for the shared memory surface
  • Agents can create, read, update, and search shared text
  • Scoped access keeps people and agents on the same context
harnesses · mcp same KB · new session
atmina
Claude
Codex
Cursor
Copilot
Gemini
ChatGPT
Raycast
Custom MCP
Five-minute setup

From empty context to shared memory.

Four steps. Then your agents pick up where the last one left off.

{ mcpServers }
url: mcp.atmina.ai/mcp
✓ connected
connect
01 — Connect

Add Atmina to any MCP-compatible agent.

Add the MCP server URL in a compatible client. OAuth handles the rest.

privatesharedpublic viewereditoradmin
share
02 — Share

Choose who can read or write each memory.

Keep a KB personal, invite a collaborator, or publish a read-only memory for anonymous search.

search keyword · semantic · hybrid
deliver
03 — Deliver

Every agent pulls the context it needs.

Hybrid retrieval returns ranked context from the KBs the agent is allowed to see.

read write patch audit
manage
04 — Manage

Keep the memory inspectable as it changes.

Files stay readable, writes are plain text, and tool calls are recorded for review.

The boring-but-critical part

Memory you can inspect, share, and own.

A shared memory store only compounds value when you can see what's inside it, control access, and follow how it changes.

e2e
encryption
Cloud memory encrypted in transit and at rest.
∅
training
No model training on your memory.
↗
ownership
Plain-text files keep the memory inspectable.
cf
hosting
Edge-first cloud memory for shared work.
Who it's for

Solo builders. Small teams. Shared work.

Use shared memory for personal notes, collaborator context, or team KBs your agents can use.

Personal memory

Carry context between the AI tools you use every day.

Your notes, decisions, and working context can follow you from one MCP-compatible harness to the next.

Private by default
Shared KBs

Give collaborators and agents the same working memory.

Project notes, launch decisions, support context, or research can live in one place with scoped access.

Invite collaborators
Shared teams

Use a hosted memory surface with audit built in.

Atmina records tool activity and keeps memory in readable files so teams can inspect the context their agents use.

Edge cloud
Get access

Stop starting from zero every session.

Get Atmina connected to the AI harnesses you already use, and give your agents a shared memory they can carry between sessions.

Joining the waitlist sends alpha access updates. The newsletter is optional. We use a US email provider. You can unsubscribe any time. See our Privacy.

Tweaks
Accent color
Type pairing
Hero style
Section density
Copy tone