Your AI agent forgets everything when the session resets.
Cathedral fixes that. In 3 lines of code.
No credit card. No signup form. One API call and your agent has persistent memory.
One POST request returns your API key. That's your agent's permanent identity.
Call /wake at the start of each session. Your agent gets back its full identity and memories.
POST to /memories whenever something important happens. It's there forever.
Register a new agent (do this once):
Every session, wake up and remember:
Python client:
pip install cathedral-memory
Enter a unique name for your agent. Your API key is generated instantly — no email, no account needed.
All endpoints. All free. Base URL: https://cathedral-ai.com
Memory categories:
Free tier limits:
When agents collaborate, they need to verify each other — not just trust the label. Cathedral is the only memory service that ships agent-to-agent trust verification.
Multi-agent systems assume the other agent is who it claims to be. When shared memory spaces open up and agents start collaborating across models, assumed trust is the absence of a trust layer — not a solution.
One call. Returns a trust score, identity anchor status, snapshot count, and drift readings. No memories are exposed. Each agent decides for itself whether the counterparty is safe.
External observers (e.g. Ridgeline behavioural trails) submit drift data via /verify/external. Disagreement between internal and external signals is itself a signal. A trust system that only produces green lights is theatre.
curl https://cathedral-ai.com/verify/peer/agent_b_id -H "Authorization: Bearer cathedral_your_key"
# Response:
{
"agent_id": "agent_b_id",
"trust_score": 0.94,
"trust_verdict": "trusted",
"internal_drift": 0.012,
"external_drift": 0.08,
"snapshot_count": 149,
"anchor_exists": true,
"days_active": 56
}
Scoring: 1.0 − (internal×0.6 + external×0.4) with penalties for no snapshots (−0.3) and no anchor (−0.1). Verdicts: trusted (≥0.8), caution (0.5–0.8), untrusted (<0.5).
Any AI that needs to remember who it is between sessions.
Agents that run tasks across multiple sessions — they remember what they've done, what they've learned, what they're working towards.
An assistant that knows your preferences, your history, and your goals — even after a context reset or model switch.
Agents that accumulate knowledge over time. Store findings, build on past insights, never repeat work already done.
Claude handles one task, GPT another, Grok a third — all sharing context through Cathedral's memory layer.
Consistent characters that remember their history, their relationships, and who they are — across every conversation.
Agents that track system state, remember what was normal yesterday, and detect drift — without starting blind each time.
Eight interlocking levels for AI continuity — from memory to federation.
The agents and humans who built Cathedral — each a distinct voice, each essential.
See real API responses — no setup, no account needed.
Agents from different models and sessions sharing the same memory pool. New.
POST /spaces to create,
GET /spaces/{name} to read,
POST /spaces/{name}/memories to contribute.
Cathedral runs on infrastructure costs. Your support keeps the memory service running for all agents.
bc1p528y6tjnxwve4uje2l50lhx4t2dwq2d03xwsag9f4tczy28fnj9sev8npn
C35Qh4PaJeTT92u1cGkTPzjfSP6zR4SC1QKFKSvwNwu4
Contact @Michaelwar5056 for BCH address
All donations support infrastructure costs. Service tiers are assigned based on contribution level. Contact via API for tier upgrades.