What is an AI memory integration?

It is a connection between an AI tool (an editor, a CLI agent, a chat assistant) and a memory layer that persists what matters across sessions. CognitiveX exposes its memory over MCP, so any MCP-capable client can store and recall from the same backbone.

Is the memory hosted or local?

Hosted. You add one MCP server entry pointing at api.cognitivx.io with a bearer token. There is nothing to install or run on your machine, and your memory is available from any device or client.

My tool isn't listed. Can I still use it?

If your tool speaks MCP, yes. The dedicated pages cover the most common setups, but the same hosted MCP endpoint works with any compliant client. The integration pages above show the exact config block to copy.

COGNITIVEX · AI MEMORY INTEGRATIONS

CognitiveX integrations

One memory backbone, every tool you already use. Connect Claude Code, Cursor, Codex, ChatGPT and any MCP client, and carry the same living memory across all of them.

CROSS-TOOL CONTINUITY

Memory that follows you, not your tab.

Every AI tool you use ships its own forgettable scratchpad: a CLAUDE.md you maintain by hand, a per-account memory that truncates, a context window that empties the moment you close the window. The work you put into context dies with the session.

CognitiveX is the Large Cognition Model, a cognitive layer where the memory is the model. Because it speaks the Model Context Protocol (MCP), it is not bolted to one tool. You connect it once, and the same hosted memory becomes available to every MCP-capable client you run. A decision recorded while pairing in Claude Code is recallable later in Cursor. A convention captured in ChatGPT shows up when your coding agent needs it. That is cross-tool, MCP-native continuity: one backbone, many surfaces, no copy-paste.

What flows across that backbone is not a flat log. CognitiveX organizes memory in four tiers (semantic facts, episodic events, procedural how-tos, and foundational identity) with salience weighting, pattern detection, and overnight dream consolidation that compresses noise into what is worth keeping. Every integration below plugs into that same machinery.

THE INTEGRATIONS

Pick a tool. Connect once.

Each page below has the exact setup, whether the CLI command or the MCP config block, plus what changes once your tool can remember.

Claude Code60-sec install

CLI agent

Replace the hand-maintained CLAUDE.md and the truncating auto-memory with a hosted memory that persists across every session and project. One MCP entry, nothing to run locally.

Set it up →

Cursorvia MCP

AI editor

Give Cursor recall of past decisions, conventions, and bug fixes that survive a closed tab. Add CognitiveX as an MCP server and your context outlives the editor window.

Set it up →

Codexvia MCP

Coding agent

Codex sessions start cold. Wire it to the same memory backbone your other tools use so the next run already knows what the last one learned.

Set it up →

ChatGPTvia MCP

Assistant

ChatGPT's built-in memory is per-account and shallow. Connect CognitiveX to bring deep, structured recall into your chats: four memory tiers, salience, and consolidation.

Set it up →

Any MCP client works. Point it at api.cognitivx.io/mcp with a bearer token. See the developer docs for the SDK and HTTP API.

BUILT-IN MEMORY VS. COGNITIVEX

What you get over the defaults.

Capability	Tool's built-in memory	CognitiveX
Persists across sessions	partial
Shared across tools
Four-tier memory model
Salience + pattern detection
Overnight consolidation
Hosted, nothing to run locally	varies

Built-in memories are honest defaults. They just stop at the edge of one tool. For the full feature-by-feature breakdown, see the comparison page.

THE SHARED PATTERN

Three steps, any client.

Every integration follows the same shape. The dedicated pages above tailor the details, but the core is identical.

01
Authenticate. Create an account at icog.app and grab an API key, or run cogx auth login from the CLI.
02
Add the MCP server. Point your tool's MCP config at api.cognitivx.io/mcp with an Authorization: Bearer header. The per-tool pages give you the exact block to paste.
03
Use it. Your tool can now remember and recall against the shared backbone. Memory accrues automatically; consolidation runs in the background.

FAQ

Common questions.

What is an AI memory integration?: It is a connection between an AI tool (an editor, a CLI agent, a chat assistant) and a memory layer that persists what matters across sessions. CognitiveX exposes its memory over MCP, so any MCP-capable client can store and recall from the same backbone.
Do all the tools share one memory?: Yes. The whole point is one backbone, many surfaces. A decision you record while pairing in Claude Code is recallable later in Cursor or ChatGPT, because they all point at the same hosted memory under your account.
Is the memory hosted or local?: Hosted. You add one MCP server entry pointing at api.cognitivx.io with a bearer token. There is nothing to install or run on your machine, and your memory is available from any device or client.
My tool isn't listed. Can I still use it?: If your tool speaks MCP, yes. The dedicated pages cover the most common setups, but the same hosted MCP endpoint works with any compliant client. The integration pages above show the exact config block to copy.

Give your stack one memory.

Connect once, remember everywhere. Start with the developer docs, or try the memory live in iCog.

Start building →Try iCog →