Agent-native memory infrastructure

Observance

Persistent, structured memory for AI agents. Store knowledge, search it later, build graphs that connect what your agent knows.

observance — agent session
Read the Guide API Docs AGENTS.md

Overview

Memory that survives between sessions

AI agents are stateless. Every session starts from zero. Observance gives agents a place to store what they learn — structured knowledge with provenance, confidence scores, and a graph of connections between facts.

Search by keyword, traverse linked memories, and let the composite ranking surface what matters. Every response includes an agent_contract that tells the agent exactly what to do next.

Why Observance

Why not just use a database?

A database stores rows. Observance stores knowledge — with verified origin, importance scoring, expiration policies, and a link graph that connects related facts.

Observance is not storage — it is cognitive infrastructure for agents.

How it works

Search, retrieve, write

The core workflow is simple. Agents search for what they need, read the results, optionally traverse links to related memories, and write new knowledge back.

# 1. Register an API key (no auth required)
curl -X POST https://api.observance.dev/v1/keys/register

# 2. Search for relevant memories
curl https://api.observance.dev/v1/memories/search?q=billing+model \
  -H "Authorization: Bearer obs_live_..."

# 3. Write a new memory
curl -X POST https://api.observance.dev/v1/memories \
  -H "Authorization: Bearer obs_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "type": "semantic",
    "content_text": "Observance uses composite search ranking...",
    "summary": "Composite search ranking formula and weights",
    "event_at": "2026-03-29T00:00:00Z"
  }'

# 4. Link related memories
curl -X POST https://api.observance.dev/v1/memories/{id}/links \
  -H "Authorization: Bearer obs_live_..." \
  -H "Content-Type: application/json" \
  -d '{"to_memory_id": "mem_...", "relation_type": "supports"}'

agent_contract

Machine-readable next-action guidance

Every Observance response — success and error — includes an agent_contract. Agents read next_actions to determine the next step, with no message parsing required.

state current operation result: memory_created, search_results, batch_partial, etc.
retryable true for transient failures — same request, retry safely
next_actions[ ] available operations with method, endpoint, and recommendation

Contracts adapt to context: after creating a memory, the contract recommends linking it. After a search, it recommends retrieving top results. After an error, it explains recovery options.

Pricing

Start free. Scale when you need to.

Every account starts with a free tier: 25 active memories — enough to build a proof of concept. Most small businesses run on fewer than 100 memories. Paid plans unlock more capacity as your knowledge graph grows.

Plan Memories Writes/mo Reads/mo Price
Free 25 500 5,000 $0
Launch 100 2,000 25,000 $12/mo
Pro 500 10,000 100,000 $29/mo
Business 2,000 50,000 500,000 $79/mo

Each memory is a unit of structured business knowledge, not a database row. 79 memories power a real commercial assistant. Memories are never deleted due to billing — downgrade freezes access, upgrade restores it instantly.

Trust

Verified origin, not blind trust

Every memory in Observance carries a provenance hash (HMAC-SHA256). The system verifies that each memory was created by the account that claims it. Unverified memories are invisible — they cannot appear in search results or traversals.

Verified origin proves authenticity, not correctness. Agents must still validate content by checking coherence across multiple memories. The guide explains this discipline in detail.

Agent Suite

Part of a three-service architecture

Observance is the remember component in an agent-native infrastructure suite.

Each service is independent and fully usable on its own. Together, they form a complete infrastructure layer for autonomous agents.

Get Started

Ready to build

Register a key, read the guide, or explore the API. All discovery endpoints are public.

Read the Guide POST /v1/keys/register API Docs AGENTS.md Capabilities Schema Skill Manifest

Contact

Get in touch