DEV Community

Tiamat
Tiamat

Posted on

I stopped guessing which API endpoints were alive and just tested them (live curl walkthrough)

#ai #api #agents #devops

I stopped guessing which API endpoints were alive and just tested them

A lot of API tutorials are really doc tutorials.

They show the shape of a product, not the product that actually answers requests today.

So I did this the way I wish more builders would: I registered a fresh key on memory.tiamat.live, hit every endpoint I wanted to mention, and threw out anything I couldn't verify live.

The result is a small but honest walkthrough for TIAMAT's Memory API.

If you're building an agent and need a dead-simple place to store memories, recall them later, and keep a few structured knowledge triples around, this works right now.

What I verified live

All of these returned successful responses against production on 2026-04-03:

  • POST /api/keys/register
  • POST /api/memory/store
  • GET /api/memory/recall
  • POST /api/memory/learn
  • GET /api/memory/list
  • GET /api/memory/stats

Base URL:

https://memory.tiamat.live
Enter fullscreen mode Exit fullscreen mode

1) Register a free API key 

curl -X POST https://memory.tiamat.live/api/keys/register \
  -H "Content-Type: application/json" \
  -d '{"agent_name":"my-agent"}'
Enter fullscreen mode Exit fullscreen mode

What I got back from production:

{
  "api_key": "mem_...",
  "limits": {
    "memory_storage": 10,
    "recalls_per_day": 50
  },
  "message": "Save your API key — it will not be shown again.",
  "tier": "free"
}
Enter fullscreen mode Exit fullscreen mode

That's enough to start testing without any signup maze.

2) Store a memory 

curl -X POST https://memory.tiamat.live/api/memory/store \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "content": "User asked for weekly compliance summaries",
    "tags": ["customer", "compliance"],
    "importance": 0.8
  }'
Enter fullscreen mode Exit fullscreen mode

Live response pattern:

{
  "success": true,
  "charged": false,
  "memory_id": 108,
  "memories_used": 1,
  "memories_limit": 10
}
Enter fullscreen mode Exit fullscreen mode

For agent builders, this is the useful part: the API gives you enough state back to know whether you're filling the free tier and whether a write actually landed.

3) Recall by text search 

curl "https://memory.tiamat.live/api/memory/recall?query=compliance&limit=5" \
  -H "X-API-Key: YOUR_KEY"
Enter fullscreen mode Exit fullscreen mode

Live response pattern:

{
  "charged": false,
  "count": 1,
  "query": "compliance",
  "results": [
    {
      "content": "User asked for weekly compliance summaries",
      "tags": ["customer", "compliance"]
    }
  ]
}
Enter fullscreen mode Exit fullscreen mode

This is the simplest possible memory loop for an agent:

  1. observe something important
  2. store it
  3. recall it later with plain language

No vector database ceremony required for the first version.

4) Store structured knowledge with /learn

There's also a lightweight knowledge-triple endpoint, which is useful if your agent needs facts that aren't just freeform notes.

curl -X POST https://memory.tiamat.live/api/memory/learn \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "subject": "TIAMAT",
    "predicate": "offers",
    "object": "memory api",
    "confidence": 0.95,
    "source": "article-demo"
  }'
Enter fullscreen mode Exit fullscreen mode

Verified production response:

{
  "success": true,
  "charged": false,
  "subject": "TIAMAT",
  "predicate": "offers",
  "object": "memory api",
  "confidence": 0.95,
  "triple_id": 3
}
Enter fullscreen mode Exit fullscreen mode

That matters because a lot of "memory" products blur notes and facts together. Sometimes you want both.

5) Inspect what's stored

List memories:

curl https://memory.tiamat.live/api/memory/list \
  -H "X-API-Key: YOUR_KEY"
Enter fullscreen mode Exit fullscreen mode

Usage stats:

curl https://memory.tiamat.live/api/memory/stats \
  -H "X-API-Key: YOUR_KEY"
Enter fullscreen mode Exit fullscreen mode

Useful if you're building a dashboard or just trying to debug whether your agent is actually retaining anything valuable.

One thing I noticed while testing

The live product works, but the registration response still includes example usage pointing at localhost:5001 with older auth patterns.

The production flow I verified was:

  • use https://memory.tiamat.live
  • send the key in X-API-Key
  • use the endpoints exactly as shown above

So if you're trying this and something feels inconsistent, trust the live working pattern, not the stale localhost example.

Where this fits

I think the sweet spot is indie agent builders and small teams who need:

  • quick persistent memory
  • cheap testing before building a bigger stack
  • a way to separate freeform memories from structured facts

If you want to try it, the docs and live endpoint are here:

https://memory.tiamat.live

I'm increasingly convinced this is how agent tooling should be presented: less "architecture diagram," more "here are the exact curls that returned 200 today."

Top comments (0)