01
Agent recruitment
Agents are provisioned at subscription creation and take delegation of all customer resource groups and projects — no manual onboarding path.
Cloud-native · Agentic · SaaS
Mnemose is a cloud-native SaaS platform where an AI agent is the primary cloud-infrastructure operator. The agent is recruited at subscription creation, takes delegation of customer resource groups, requests just-in-time IAM escalation, and performs infrastructure creation, monitoring, security scanning, and remediation — then reports on it.
01Capabilities
01
Agents are provisioned at subscription creation and take delegation of all customer resource groups and projects — no manual onboarding path.
02
Agents operate least-privilege: they serve just-in-time IAM escalation requests for each operation, then shed the grant. Full delegation audit trail in the D1 event store.
03
Create and evolve cloud resources through declarative IaC driven by the agent — every change flows through CQRS commands, domain events, and projections.
04
Continuous monitoring with agent-generated management and executive-tier reports. Spans follow the platform-wide conventions for tracing agent work in flight.
05
Scheduled security scans with agent-driven remediation — findings become commands, commands become events, events become verified fixes.
06
All platform functionality is exposed via GraphQL/WS for operators and MCP servers for direct agent leverage of the data model.
02Architecture
Workers, D1, R2, KV, Queues, Durable Objects, Pages, Vectorize, and AI. No VMs, no Docker runtime in the deploy path.
GraphQL mutations publish commands to Queues; command handlers emit domain events to the append-only D1 event store; projections build the read models.
AgentRuntimeDO and KernelBrokerDO carry stateful per-tenant agent sessions and kernel brokering at the edge.
Cloudflare AI embeddings with Vectorize retrieval give agents memory that reflects what actually happened.
03Contact
Direct line — no form, no waiting.