Early access · In development · MCP native

Your AI finally
remembers you.

Every conversation starts from zero. You re-explain your project, your preferences, your context. Echo is a memory engine for AI agents. We host it for you, or deploy it for your team. It's in active development and available by early access.

Request early access See hosted & team options

Need help getting started? Email Devontae for access →

Works with Claude, GPT, Gemini, Cursor, and any MCP client.

10
MCP Tools
12+
REST Endpoints
3
Search Modes
<50ms
Search Latency
60%

of AI tokens are wasted re-reading context your AI should already know

Every time

you start a new chat, your AI forgets everything you've ever told it

Scattered

notes in 5 apps, voice memos you never listen to, decisions you can't find

Built for two kinds of people

People who use AI every day

"I told Claude about my project last week and now it has no idea." Echo gives your AI a memory so every conversation starts with context, not from scratch.

Developers building with AI

Add persistent memory to any app in five lines of Python. Your chatbot remembers users. Your agent remembers decisions. Your pipeline keeps context across runs.

example.py 
import requests

# Self-host: API = "http://localhost:3002/api/v1"
# Hosted:    API = "https://echo-memory.fly.dev/api/v1"
API = "http://localhost:3002/api/v1"
headers = {"X-API-Key": "echo_sk_..."}

# Store a memory
requests.post(f"{API}/memories", json={
    "content": "Q2 launch needs design review by Friday",
    "tags": ["launch", "q2"],
    "domain": "work"
}, headers=headers)

# Search — hybrid (FTS + vector)
r = requests.get(f"{API}/memories/search",
    params={"query": "Q2 design deadline"},
    headers=headers)
print(r.json()["data"])
# [{content: "Q2 launch needs design review...", score: 0.94}]

How it works

Store anything. Find it instantly. Your knowledge compounds over time.

1. Capture

Save notes, decisions, meeting takeaways, research, anything. Tag it or don't. Echo indexes everything for search automatically.

2. Search

Find what you need by meaning, not just keywords. "What did we decide about the launch?" works even if you never typed the word "launch."

3. Compound

Memories link to each other. A note from January connects to a decision from yesterday. Your second brain grows with you.

Under the hood

Full architecture details →

Three ways to run Echo

Same engine underneath. You choose how much of the work you want to do, and how much you want us to do for you. Echo is in active development; early access is available by request.

EARLY ACCESS

Self-Host

Requestaccess

Run the binary on your own hardware and own everything. Available to early-access users.

  • ✓ Run it on your own infrastructure
  • ✓ Runs on Mac, Windows, Linux, Raspberry Pi
  • ✓ Full API: search, store, graph, tasks
  • ✓ MCP native + REST API
  • ✓ 100% local. Zero cloud. Your data stays yours.
  • · Direct support from the builder
Request early access

Questions first? Email Devontae →

WE RUN IT

Hosted

$99/mo

For people who want Echo without operating it. We handle uptime, backups, updates.

  • ✓ Managed instance, single-tenant
  • ✓ Daily backups, monitored 24/7
  • ✓ Free version updates as they ship
  • ✓ MCP native + REST API + examples
  • ✓ Email support, <48hr response
  • ✓ Cancel anytime, export your data
Start Hosted
FOR TEAMS

Deployed by Us

Let's talk

For teams who want Echo running on their infrastructure, integrated with their stack.

  • ✓ On-prem or your cloud account
  • ✓ SSO, audit logs, role-based access
  • ✓ Custom integrations with your tools
  • ✓ Team training included
  • ✓ Priority support & SLA
  • ✓ Delivered by Textstone Labs
Talk to Textstone

Not sure which fits? Most individuals self-host. Most small teams start hosted. Anything bigger usually wants us to deploy it. Ask us →

See all features → View architecture →

Built because I needed it

Echo isn't a startup idea. It's infrastructure I built for myself and couldn't stop using.

1,000+
memories stored
10
MCP tools in production
24/7
running on a $60 Raspberry Pi

I'm a consultant, MBA student, and founder building three products at once. Every Claude Code session started from zero. I was re-explaining my architecture, re-discovering dead ends, re-making decisions I'd already made.

So I built Echo. A Go binary on a Raspberry Pi that remembers everything across every session, every project, every tool. Voice notes get transcribed and ingested while I sleep. Decisions get linked into a knowledge graph. Every morning, my AI already knows what happened yesterday.

After two months and 1,000+ memories, I realized this was the most valuable thing I'd built. Now it's yours.

Dogfooded daily since January 2026

Echo wasn't built to ship. It was built to use. These are real numbers from our own production instance.

1,000+
memories stored
451
journal entries auto-generated
7
projects tracked simultaneously
25
security fixes before first beta invite

We run Echo across consulting engagements, MBA coursework, PhD research, and three product builds. Voice notes get transcribed and ingested while we sleep. Decisions link into a knowledge graph. Every morning, the AI already knows what happened yesterday.

When we have real beta tester feedback, it'll go here. Until then, we'd rather show you real numbers than fake quotes.

Ready to give your AI a memory?

In active development · Early access by request · Five minutes to first API call.

Request early access Start Hosted ($99/mo)

Want us to deploy it for your team? Talk to Textstone Labs. We handle on-prem, integrations, training, and SLA.