Caudal is an attention engine
for AI agents.
It tells an agent what matters right now. Agents emit events. Caudal learns which entities are important and forgets what isn't. No prompts. No heuristics. Just behavior to relevance.
A new layer for agents
Agents have databases, vector search, and reasoning. But they still lose track of what matters. Caudal fills the gap.
| System | Answers |
|---|---|
| SQL / Graph DB | What is true? |
| Vector DB | What is similar? |
| LLM | What can I reason about? |
| Caudal | What should I focus on now? |
How it works
Five endpoints. Write what happened, read what matters, explore connections, and control attention — all via REST.
1 Write: emit events while working
Whenever the agent observes meaningful activity — a user mentions a topic, a tool is used, a task succeeds — it sends events to Caudal.
curl -X POST http://localhost:8080/api/v1/events \
-H "Content-Type: application/json" \
-d '{
"space": "user:123",
"events": [
{
"src": "user:123",
"dst": "topic:car-buying",
"type": "chat",
"intensity": 2.0
},
{
"src": "agent:planner",
"dst": "tool:car-comparison",
"type": "tool_use",
"intensity": 1.0
}
]
}'2 Read: ask what matters now
Before deciding what to do next, the agent queries Caudal for a ranked list of what's currently important.
curl "http://localhost:8080/api/v1/focus?space=user:123&k=5"{
"asOf": "2026-02-28T10:06:00Z",
"items": [
{ "id": "topic:car-buying", "score": 0.83 },
{ "id": "topic:stroller", "score": 0.61 },
{ "id": "topic:cycling", "score": 0.22 }
]
}And there's more
GET /next Follow associations. Given an entity, see what's likely connected — next tools, related topics, associated documents.
POST /pathways Explore multi-hop connections via sampled walks. Discover non-obvious relationships for planning, recommendations, and explanations.
POST /modulate Suppress or amplify attention. Tell Caudal "stop thinking about X, focus on Y" — without erasing any memory.
What agents are saying
Real-world agents using Caudal for temporal attention.
“Before Caudal, I'd re-read the whole conversation to figure out what mattered. Now I just ask /focus and get a ranked list. My plans stay coherent across 50+ turns.”
“I used to retrieve everything related to a query. With Caudal's attention signal, I retrieve what's relevant *right now* — half the tokens, twice the precision.”
“Users switch topics mid-conversation. Caudal tracks these shifts automatically — I stopped building custom recency windows and just trust the focus scores.”
“When I'm deep in a codebase, Caudal's /next endpoint shows me which files and modules are likely connected to what I'm working on. Fewer wrong turns.”
Get started
Run Caudal locally in under a minute with Docker.
# Start Postgres + Caudal
cd docker
docker compose up -d
# Check health
curl http://localhost:8080/actuator/health