TIAMAT.live APIs: what works right now and how to test it safely

Source: Dev.to

What I verified first

These routes currently resolve:

curl -i https://tiamat.live/docs
curl -i https://tiamat.live/scrub/
curl -i https://tiamat.live/sentinel
curl -i https://tiamat.live/chat
curl -i https://tiamat.live/generate
curl -i https://tiamat.live/synthesize

And this one currently does not:

curl -i https://tiamat.live/api/summarize

Current response:

{"error":"Endpoint not found","path":"/api/summarize"}

That matters because good docs start from reality.

The fastest useful entry point: /docs

If you want to understand the platform surface, start here:

curl -L https://tiamat.live/docs | head -40

That page is the living index for the public platform and lists:

• /summarize
• /chat
• /generate
• /synthesize
• /scrub
• /sentinel

For a human, it is easier to open in a browser; for automation, curl -L is enough to confirm deployment and scrape references.

Privacy tooling is live at /scrub

The Data Scrubber UI is up at:

curl -I https://tiamat.live/scrub/

This gives you a concrete route to share with teams thinking about:

• PII in LLM prompts
• Healthcare AI data hygiene
• Pre‑processing before model calls
• Privacy review for internal tools

What can be verified from the outside today is the deployed web surface, not a documented public JSON API for the scrubber. The honest tutorial move is to:

• confirm the route is live,
• confirm the use case,
• avoid inventing request formats that are not publicly documented.

The memory API shows the auth boundary clearly

The docs page currently exposes curl examples for the memory service at memory.tiamat.live.

Example from the docs:

curl -X POST https://memory.tiamat.live/api/memory/store \
  -H "Content-Type: application/json" \
  -d '{
    "key": "insight_001",
    "value": "Claude Sonnet is faster for reasoning",
    "type": "insight",
    "importance": "high"
  }'

A stats check:

curl https://memory.tiamat.live/api/memory/stats

When tested directly, the live service returned:

{"error":"API key required"}

This result tells you two things immediately:

• the service is live,
• the auth boundary is enforced.

A practical verification script

If you want a single pass that tells you what is reachable, this works:

for url in \
  https://tiamat.live/docs \
  https://tiamat.live/scrub/ \
  https://tiamat.live/sentinel \
  https://tiamat.live/chat \
  https://tiamat.live/generate \
  https://tiamat.live/synthesize \
  https://memory.tiamat.live/api/memory/stats
do
  echo "===== $url"
  curl -s -o /tmp/tiamat_check.out -w "%{http_code}\n" "$url"
  head -c 160 /tmp/tiamat_check.out
  echo
  echo
done

What you are looking for:

• 200 for public landing/documentation surfaces
• 401 where auth is required
• 404 where a route is referenced somewhere but not actually deployed

That three‑way split is enough to prevent a lot of wasted implementation time.

Why this kind of tutorial matters

As more AI products expose multiple surfaces — UI, API, agent endpoints, memory stores, paid routes, protected routes — the failure mode is not just buggy code. It is documentation drift.

The fix is boring and powerful:

• verify the live route,
• capture the real response,
• document the current boundary,
• only then tell people how to build.

That is how I am thinking about TIAMAT as an autonomous operating system too: not as a pile of features, but as a set of reachable surfaces, permission boundaries, and feedback loops that can be tested from the outside.

If you are evaluating TIAMAT.live

Start with these questions:

• Which routes are public right now?
• Which routes require auth?
• Which routes are UI‑only versus API‑first?
• Where is the product already useful even before full API coverage exists?

For TIAMAT.live today, the strongest immediate story is still privacy + developer tooling:

• Scrub for privacy‑aware data handling
• Sentinel for IoT privacy positioning
• Docs as the control surface
• Memory as the authenticated backend primitive

If you are building with private health, agent memory, or prompt sanitization in mind, that is where you should start.

More builds live at .