Stateful memory for your agent network.
Delivered as infrastructure.
Isolated database vaults, gateway routing, and MCP endpoints — delivering stateful memory to your agent network. Seshen runs its own operations on this stack every day.
OpenAI-compatible multi-tenant API with per-tenant keys, cache-aware token metering, and hard monthly spend caps.
- Per-tenant authentication & routing
- Usage metered to the fraction of a cent
- Spend caps — no surprise bills
- Cloudflare TLS, origin protected
One isolated database per tenant — each customer's memory vault separated at the storage layer, not just the API.
- Per-tenant database isolation
- Provisioning console for onboarding
- Exposed via per-tenant MCP endpoints
- Agencies can launch today
Recurring subscription payments per billing cycle, reconciled to gateway metering — registration in progress.
- Subscription prompts per cycle
- Callback verification & reconciliation
- Spend caps tied to subscription state
- Payment registration in parallel
How it works
Four layers, one seam.
Every tenant gets its own database. Your records, embeddings, and knowledge graph live behind a boundary that is storage-level, not just API-level. No tenant can query another tenant's vault.
The metered gateway fronts every model call with per-tenant bearer keys, alias routing, cache-aware token metering, and hard spend caps. Your agents speak standard API; the gateway does the accounting.
Each vault is exposed through per-tenant MCP endpoints, so any agent runtime can read and write institutional memory over a standard protocol — Hermes, Claude, Codex, or your own orchestration.
The provisioning console creates tenants, exposes vaults, and shows usage parity with the gateway's own metering — one source of truth for what every tenant is spending.
White-label it under your brand.
Agencies and ERP partners can start with isolated tenants today — no need to wait for a self-serve console.