MemoryRelay.ai
ICM workspaces · Coming soon

Versioned context for AI agents

MemoryRelay is becoming an ICM service. Build the context your agents work from as an Interpretable Context Methodology workspace — ordered stages, clear contracts, human review — and deliver exactly what each stage needs.

  • Stage-by-stage context your agents walk as plain files
  • Versioned releases, so every run knows what it was given
  • Human review at every stage, on the record
Learn about ICMJoin the waitlist

Available today: the MemoryRelay memory layer. Start free.

text
release-notes/
├── CLAUDE.md          # entry: where to go
├── CONTEXT.md         # always-true context
└── stages/
    ├── 01-collect/
    │   ├── CONTEXT.md # stage contract
    │   └── output/
    ├── 02-draft/
    │   ├── CONTEXT.md
    │   └── output/
    └── 03-review/
        ├── CONTEXT.md
        └── output/

An illustrative ICM workspace: one entry file, ordered stages, a contract per stage.

What is ICM?

The Interpretable Context Methodology organises what an agent needs to know as a small workspace of plain files it can walk, instead of a pile of context it has to search.

Folders carry sequencing

Work is split into ordered stages. Each stage is a folder with a contract that says what it takes in, what it produces, and who checks it.

Hierarchy scopes context

An agent walks from one short entry file to the stage it needs, and loads only what that stage lists — not everything you have ever written down.

Plain files carry state

Context, decisions and outputs are ordinary files. A person can read them, review them and keep them in git; an agent can walk them with no service running.

How MemoryRelay builds on it
Available today

The memory layer

Persistent memory, semantic search and knowledge graphs for your agents, over REST, MCP and the Python SDK. In ICM it becomes the evidence your workspaces draw on.

Memory Powers Every AI Use Case

From personal assistants to enterprise automation, persistent memory transforms what AI agents can accomplish.

AI Assistants

Give your chatbot memory that persists across sessions. Users never have to repeat themselves.

Customer Support

Agents that remember past tickets, preferences, and solutions. Faster resolution, happier customers.

Education & Tutoring

Track learning progress, adapt to student knowledge, and build on previous sessions automatically.

Healthcare Agents

Maintain patient context securely across visits. Enable continuity of care with AI that remembers.

E-Commerce

Product recommendations that improve over time. Agents that understand purchase history and taste.

Enterprise Workflows

Automate multi-step processes where agents need context from previous interactions and decisions.

Built for Production

Everything you need to give your AI agents persistent memory. Reliable, fast, and easy to integrate.

Blazing Fast
Fast semantic search powered by pgvector. Store and retrieve memories with minimal latency.
Cost Efficient
Flat-rate pricing with a generous free tier. No per-operation charges. Self-host for free forever.
Developer First
Python SDK, MCP server, REST API. Comprehensive docs and examples. Integrate in minutes.
Production Ready
Encryption at rest and in transit. API key scoping. Rate limiting. Built for developers who ship.

Get Started in Minutes

Three simple steps to give your AI agents persistent memory.

STEP 01

Install the SDK

One command to get started. Available for Python with more languages coming soon.

bash
pip install memoryrelay
STEP 02

Store Memories

Save context from conversations, interactions, and decisions. Embeddings are generated automatically.

python
from memoryrelay import MemoryRelay
client = MemoryRelay(api_key="mem_sk_...")
client.memories.create(
    content="User prefers dark mode",
    agent_id="my-agent"
)
STEP 03

Search & Retrieve

Semantic search finds relevant context even with different wording. Build smarter agents instantly.

python
results = client.memories.search(
    query="What does the user prefer?",
    agent_id="my-agent"
)
# Returns: "User prefers dark mode" (score: 0.92)

Works With Your Stack

Official Python SDK, REST API, and MCP server. Integrate with any language or framework.

python
1from memoryrelay import MemoryRelay
2
3client = MemoryRelay(api_key="mem_sk_...")
4
5# Store memories
6client.memories.create(
7    content="User prefers dark mode and Python",
8    agent_id="assistant",
9    metadata={"source": "chat", "importance": "high"}
10)
11
12# Semantic search
13results = client.memories.search(
14    query="What are the user's preferences?",
15    agent_id="assistant",
16    limit=5
17)
18
19for result in results.data:
20    print(f"Score: {result.score:.2f} - {result.memory.content}")

Start Building with MemoryRelay

Free to start with a 14-day Builder trial. No credit card required. Get your AI agents remembering in under 5 minutes.

Create Free AccountView Pricing