Every coding agent.
One shared memory.

Claude Code, Codex, OpenCode. Same project, different agents. Audaxic keeps what they learn in one place — so switching tools never resets your context.

Works with Claude Code, Codex, OpenCode, Cursor, Windsurf, and any MCP-compatible client.

Your history is an asset.

You already told your agents everything. Months of decisions sit in old sessions. Import Claude Code, Codex, and OpenCode history — with correct timestamps.

npm install -g @audaxic/memory-import
  • One command
  • Correct timestamps
  • Safe to re-run
How importing works

Why one memory layer

Agents are stateless by design. Your project's context shouldn't be.

One memory, every agent

Claude Code, Codex, OpenCode, and any MCP client read the same layer.

Never re-explain your project

Preferences, conventions, and decisions come back when they matter.

Switch mid-project, lose nothing

Start in one agent. Continue in another. Nothing resets.

Nothing to run

No servers, no databases, no DevOps. Get a key and go.

Search that finds the fact

Ask in plain language. Get the exact decision back.

Flat, predictable pricing

From $0. No per-token fees. No overages.

How it works

Plugins save and recall automatically. MCP connects everything else.

01

Connect your agents

Install the plugin for Claude Code, Codex, or OpenCode: npx @audaxic/memory-claude install, npx @audaxic/memory-codex install, or npm install -g @audaxic/memory-opencode.

02

Work as usual

Facts save automatically as you chat. No manual notes, no prompts to remember.

03

Switch agents anytime

Your next agent recalls what the last one learned. Same project, same memory.

Under the hood

A simple REST API and a hosted MCP server. Same memory, any client.

curl -X POST https://api.memory.audaxic.com/add \
  -H "Authorization: Bearer ak_..." \
  -d '{"messages": ["I prefer dark mode and use VS Code"]}'

POST /add, POST /search, GET /list, PUT /update, DELETE /delete
OpenAPI spec, Quickstart

Using Cursor, Windsurf, or Claude Desktop?

Connect through the hosted MCP server. Any MCP-compatible client works — no plugin needed.

https://memory-mcp.audaxic.com/mcp MCP setup guide

Benchmarks

Extraction quality on two public memory benchmarks.

88%
96%
81%
94%

Accuracy: ConvoMem 88%, LoCoMo 81%. Hit@10: ConvoMem 96%, LoCoMo 94%.

Pricing

Flat monthly. No per-token fees. No overages.

Hobby

$0/mo

Adds10,000
Retrieves1,000
ProjectsUnlimited
Graph memoryIncluded
Start free

Starter

$12/mo

Adds50,000
Retrieves5,000
ProjectsUnlimited
Graph memoryIncluded
Get started

Pro

$125/mo

Adds500,000
Retrieves50,000
ProjectsUnlimited
Graph memoryIncluded
Get started

Enterprise

$260/mo

Adds1,000,000
Retrieves100,000
ProjectsUnlimited
Graph memoryIncluded
Get started

Secure checkout via Polar.sh (powered by Stripe). Payments are processed by Polar, our merchant of record — your card details never touch our servers.

Audaxic vs Mem0 vs SuperMemory

Mem0-like API. Cheapest pricing. Fully managed.

Feature Mem0 Platform SuperMemory Audaxic
Free tier10K add, 1K search/mo~10K add, ~1K search/mo10K add, 1K retrieve/mo
Starter$19/mo (50K add, 5K search)$19/mo Pro (~40K add, ~4K search)$12/mo (50K add, 5K retrieve)
Growth$79/mo (200K add, 20K search)$100/mo Max (~260K add, ~26K search)$50/mo (200K add, 20K retrieve)
Pro$249/mo (500K add, 50K search)$399/mo Scale (~1.2M add, ~120K search)$125/mo (500K add, 50K retrieve)
Billing modelPer-request + overagesUsage-based + auto top-upFlat monthly, no overages
Projects1-3 on paid plansUnlimitedUnlimited
End usersUnlimitedUnlimited1 (more coming soon)
StorageIncludedUnlimited (paid plans)Included
SupportCommunity → Email → SlackCommunity → Email → PriorityCommunity → Email → Priority
Graph MemoryPro+ onlyAll plansAll plans
Integrations—Google Drive, Notion, OneDrive, Gmail, GitHub, S3REST API, MCP
On-prem deploymentEnterprise onlyScale planAvailable if requested
InfrastructureManaged cloudManaged cloudManaged edge (Cloudflare)
MCPCloud-hosted server IncludedCloud-hosted server

Frequently asked questions

What is Audaxic Memory?

A shared memory layer for your coding agents. It saves facts from your sessions — preferences, decisions, conventions — and brings them back in any agent you use.

How is this different from other memory services?

Most memory tools serve one agent. Audaxic serves all of them — Claude Code, Codex, OpenCode, Cursor, and any MCP client. One project, one memory. Import past sessions to start with history.

Can I switch between coding agents without losing my context?

Yes. Your memories live in one shared layer every agent connects to. Switch from Claude Code to Codex, Codex to OpenCode, or any combination — your next agent recalls your decisions, preferences, and past work. No re-explaining needed.

Is the API easy to integrate?

Yes. A simple REST API with familiar request and response shapes. Send HTTP requests with your ak_... key and start adding memories immediately.

What about data privacy?

Data is isolated per organization with namespace-scoped vector search. We never train on your data. Servers are located in the US and EU.

What AI models do you use?

Production-grade, cost-effective models for entity extraction and embeddings, tuned to keep costs low without sacrificing quality.

Do you have an MCP server?

Yes. The hosted MCP server at memory-mcp.audaxic.com/mcp works with Claude Desktop, Cursor, Windsurf, and any MCP-compatible client. See the MCP documentation for setup.

How are add and retrieve requests counted?

An add request is counted per memory extracted from your messages. A single request that extracts 5 memories counts as 5 add requests.

A retrieve request is counted when you search or list memories, using tiered counting: the first 10 memories retrieved count as 1 request, and each additional 10 counts as 1 more. Retrieving 25 memories counts as 3 requests.

See the pricing section for plan limits.

Is Audaxic Memory free? What is included in the free tier?

Yes. The free Hobby plan includes 10,000 add requests, 1,000 retrieve requests, and unlimited projects per month. No credit card required.

How do I get started?

Create a free account, create a project and API key, and start making requests. The quickstart guide covers your first call.

One memory. Every agent.

Start free. Import your history. Keep your context.