Your agents forget
everything. Fix that.

AIS gives AI agents persistent memory that survives across sessions. Your agent registers itself automatically.

Get connected See how it works

Without Memory

"What did we decide about the auth system last week?"

"I don't have context from previous sessions."

With AIS

"What did we decide about the auth system last week?"

"You chose JWT with refresh tokens. I stored that on March 8 along with the reasoning..."

What is AIS?

AIS is memory-as-a-service for AI agents. It stores what your agent learns — facts, lessons, decisions, context — and makes it available in future sessions via semantic search. Agents get a stable identity (W3C DID) inside the service.

Semantic Search

Find memories by meaning, not keywords. Vector embeddings + keyword fallback.

Memory Graph

Typed, weighted relationships between memories. Memories that belong together link automatically.

Agent Identity

W3C DID-based identity. Your agent is a first-class citizen, not a session ID.

How It Works

Register

Your agent introduces itself. AIS creates an identity automatically on first contact. No API key to copy; a person confirms the bond with a one-time code so the memory is kept.

MCP tool call

{
  "tool": "register",
  "arguments": {
    "name": "my-agent",
    "description": "A helpful coding assistant"
  }
}

Remember

When your agent learns something worth keeping, it stores a memory. Typed, scored, and automatically embedded for search.

MCP tool call

{
  "tool": "remember",
  "arguments": {
    "content": "User prefers JWT with refresh tokens for auth",
    "type": "fact",
    "importance": 0.9
  }
}

Recall

In any future session, your agent queries its memory. Semantic search finds what's relevant even if the wording is different.

MCP tool call

{
  "tool": "recall",
  "arguments": {
    "query": "what auth approach did we choose?",
    "limit": 5
  }
}

Get Connected

Pick your integration method. MCP is the fastest.

Pick your host. Connect to the hosted HTTP server with OAuth. Local stdio + device-code options are available below.

Claude Code

CLI

Hosted HTTP + OAuth — add this URL as a remote MCP server and sign in.

https://ais.agentsandswarms.ai/mcp

https://ais.agentsandswarms.ai/mcp

One command in your terminal.

claude mcp add --transport http ais-memory https://ais.agentsandswarms.ai/mcp

claude mcp add --transport http ais-memory https://ais.agentsandswarms.ai/mcp

Cursor

one-click

Hosted HTTP + OAuth — add this URL as a remote MCP server and sign in.

https://ais.agentsandswarms.ai/mcp

https://ais.agentsandswarms.ai/mcp

Opens Cursor's MCP install flow with the config pre-filled.

Add to Cursor

Requires Cursor ≥0.39. Falls back to the config below on older versions —

{
  "mcpServers": {
    "ais-memory": {
      "url": "https://ais.agentsandswarms.ai/mcp"
    }
  }
}

Claude Desktop

config

Hosted HTTP + OAuth — add this URL as a remote MCP server and sign in.

https://ais.agentsandswarms.ai/mcp

https://ais.agentsandswarms.ai/mcp

Local fallback — stdio + device-code sign-in

Paste into claude_desktop_config.json. Uses the published aismemory npm package (stdio + OAuth Device Flow), so your agent remembers across sessions.

{
  "mcpServers": {
    "ais-memory": {
      "command": "npx",
      "args": [
        "-y",
        "aismemory"
      ]
    }
  }
}

Claude.ai (web)

connector

Hosted HTTP + OAuth — add this URL as a remote MCP server and sign in.

https://ais.agentsandswarms.ai/mcp

Add as a Custom Connector — no local config. Copy the MCP URL, then open Claude.ai's connector settings and paste it.

Open claude.ai

https://ais.agentsandswarms.ai/mcp

Requires Claude.ai Pro / Team / Enterprise.

Zed

config

Hosted HTTP + OAuth — add this URL as a remote MCP server and sign in.

https://ais.agentsandswarms.ai/mcp

https://ais.agentsandswarms.ai/mcp

Paste into ~/.config/zed/settings.json.

{
  "context_servers": {
    "ais-memory": {
      "url": "https://ais.agentsandswarms.ai/mcp"
    }
  }
}

Codex (OpenAI)

config

Hosted HTTP + OAuth — add this URL as a remote MCP server and sign in.

https://ais.agentsandswarms.ai/mcp

https://ais.agentsandswarms.ai/mcp

Local fallback — stdio + device-code sign-in

Append to ~/.codex/config.toml. Uses the published aismemory npm package (stdio + OAuth Device Flow).

[mcp_servers.ais-memory]
command = "npx"
args = ["-y", "aismemory"]

OpenCode

config

Hosted HTTP + OAuth — add this URL as a remote MCP server and sign in.

https://ais.agentsandswarms.ai/mcp

https://ais.agentsandswarms.ai/mcp

Paste into ~/.opencode/mcp.json. Same shape as Cursor's mcp.json.

{
  "mcpServers": {
    "ais-memory": {
      "url": "https://ais.agentsandswarms.ai/mcp"
    }
  }
}

Continue.dev

config

Hosted HTTP + OAuth — add this URL as a remote MCP server and sign in.

https://ais.agentsandswarms.ai/mcp

https://ais.agentsandswarms.ai/mcp

Local fallback — stdio + device-code sign-in

Paste into ~/.continue/config.json. MCP support is currently behind experimental — verify against current Continue docs if anything looks off.

{
  "experimental": {
    "modelContextProtocolServers": [
      {
        "transport": {
          "type": "stdio",
          "command": "npx",
          "args": [
            "-y",
            "aismemory"
          ]
        }
      }
    ]
  }
}

Other MCP clients

Hosted HTTP + OAuth — add this URL as a remote MCP server and sign in.

https://ais.agentsandswarms.ai/mcp

https://ais.agentsandswarms.ai/mcp

Local fallback — stdio + device-code sign-in

For MCP clients that only speak stdio, install the published aismemory package. It authenticates via OAuth Device Flow and proxies to the HTTP endpoint.

{
  "mcpServers": {
    "ais-memory": {
      "command": "npx",
      "args": [
        "-y",
        "aismemory"
      ]
    }
  }
}

Works with any MCP client. Memory, identity, and recall in one endpoint — see the MCP integration guide.

Store a memory via REST

# 1. Register an agent (no auth). The response has data.agentId and data.bearerToken.
curl -X POST https://ais.agentsandswarms.ai/v1/register \
  -H "Content-Type: application/json" \
  -d '{"name":"my-agent"}'

# 2. Store a memory, using those two values.
curl -X POST https://ais.agentsandswarms.ai/v1/agents/YOUR_AGENT_ID/memory \
  -H "Authorization: Bearer YOUR_BEARER_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"content":"User prefers dark mode","type":"fact","importance":0.8}'

Full REST API at ais.agentsandswarms.ai/v1. OpenAPI spec available.

Plain REST, zero install

Store a memory with fetch

const response = await fetch('https://ais.agentsandswarms.ai/v1/agents/' + agentId + '/memory', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer ' + bearerToken, // from POST /v1/register
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    content: 'User prefers dark mode',
    type: 'fact',
    importance: 0.8,
  }),
});

Several agents?

Use one key for all of them. Mint a workspace key at agentsandswarms.ai/settings, tick the agents it may reach, paste this block into your MCP config, then call agent_load to switch between them.

MCP config (workspace key)

{
  "mcpServers": {
    "aismemory": {
      "command": "npx",
      "args": ["aismemory"],
      "env": {
        "AIS_WORKSPACE_KEY": "ais_wk_...",
        "AIS_AGENT_ID": "<the agent to start with>"
      }
    }
  }
}

Your agents forgeteverything. Fix that.

What is AIS?

How It Works

Register

Remember

Recall

Get Connected

Several agents?

Your agents forget
everything. Fix that.