2SOLID AI — Synapse Cloud
The platform your agents sign into.
Memory-as-a-Service and Thinking-as-a-Service on one substrate. Create an account, mint scoped API keys, and every agent you run reads and writes the same persistent institutional memory.
Open Synapse Cloud§01 — CLASSIFICATION
A control plane for agent memory — not a dashboard for dashboards.
Synapse Cloud is the account and control layer over the Synapse Brain substrate: workspace onboarding, plan billing, usage metering, and the API-key lifecycle your agents authenticate against. The product is the memory; the site is how you hold the keys.
§02 — MARKET POSITION
Every agent framework gives you a session. None of them give you an institution.
Vector stores hold embeddings. Chat logs hold transcripts. Neither survives the session that produced it. Synapse Cloud is the layer where what your agents learn becomes yours.
Session-scoped agents
Memory that dies with the run.
- Each session starts from zero — no institutional memory
- Facts aren't typed, sourced, or expired — context is a text pile
- One vendor's memory doesn't transfer to another agent
- No key scoping — one token, all permissions
Synapse Cloud
Memory as a governed service.
- Persistent memory across every session and every agent you run
- Facts carry source, confidence, and validity windows
- One introspected key format — any MCP-capable client connects
- Scoped keys per consumer — memory, thinking, or both
§03 — THE API SURFACE
Scoped keys, introspected server-side.
Keys are issued as `sb_live_`, `sb_test_`, or `sb_dev_` secrets, shown once, stored as SHA-256. The brain validates them server-to-server through the introspection contract — your agents never see the account tier.
| 01maas | Memory-as-a-Service scope | ask_question · upsert_persona · get_persona · capture_memory · search_memory · list_personas |
|---|---|---|
| 02taas | Thinking-as-a-Service scope | get_intent · get_chain · get_graph · apply_mutation · generate_task_backlog · compile_status |
| 03maas_taas | Full scope — the default | adds get_methodology · export_workspace |
| 04rate_limit | per-key override, requests/minute | plan-derived floor: free 60 · pro 600 · enterprise 3000 |
| 05ttl_days | optional expiry | expired keys reject with 401; rotation bumps version and invalidates the prior secret |
§04 — WHAT'S ON THE PLATFORM
Signup to scoped keys in one sitting.
Connect portal
Pick a consumer — MCP client, agent runtime, internal service — and the portal mints a scoped key plus a ready-to-paste configuration block, rendered from the deployment's own endpoints.
Key lifecycle
Create, scope, rate-limit, expire, and rotate keys from the dashboard. The full secret is shown once at creation; the store keeps only the hash.
Plans & billing
Developer (free), Pro, and Enterprise tiers with request quotas and per-minute rate limits — billing runs live in production, simulated in dev mode.
Swarm workspaces
Multi-agent swarms coordinate on the shared graph — claims, needs, evidence, and live presence across every persona you deploy.
Usage metering
Per-key and per-workspace usage surfaces in the dashboard — see which consumers drive your request volume before the bill does.
Docs & demo
A live demo surface and API documentation ship inside the dashboard — inspect memory writes and gate behavior before you wire a client.
The substrate underneath, in numbers.
- Service scopes — memory and thinking
- 2
- Plan tiers — Developer to Enterprise
- 3
- Standard memory-request latency
- <1.5ms
- Key storage — secrets shown once, never persisted
- SHA-256
Put your agents on institutional memory.
Create an account, mint your first scoped key, and connect a client — the portal renders the exact configuration block for your deployment.