ApolloSRMFor the office of the CTO
An API- and MCP-first platform — not a walled garden.
API-first by law
Every capability is a REST endpoint — 1,700+ — and a generated CI check fails the build if a route ships without its documented public surface. Nothing important lives only in the UI, so your team is never blocked waiting on us to "expose" something.
MCP-first by design
Every endpoint is also a Model Context Protocol tool — 1,300+. Claude, ChatGPT, or your own agents drive ApolloSRM the same way the UI does. Writes are default-deny and grant-gated per user, so an agent only ever does what that user is permitted to.
OpenAPI + webhooks
A generated OpenAPI 3.0 spec drops into ChatGPT Custom GPTs, LangChain, or Pydantic AI. Event subscriptions push to nine channels — including HMAC-signed webhooks — so your downstream systems react in real time instead of polling.
Secure by construction
Tenant scope, FERPA field-masking, and role-based access are injected into every query by the compiler — not left to a developer to remember. Outbound calls are SSRF-guarded, service-to-service traffic is HMAC-signed, every mutation is audited, and the platform is SOC 2-aligned.
No lock-in
Your data lives on an open API. Migration in, integration during, and a clean export out — there is no extraction project holding you hostage. Keep the specialized tools you love via adapters and LTI 1.3.
Why the AI actually works
Because every capability is a tool an agent can drive — the same rails our own UI runs on — the AI crew can finish real work on the record, not just draft messages. The platform and the agents share one surface.
Drive ApolloSRM from ChatGPT
Point a Custom GPT at our OpenAPI spec, or connect any MCP client, and you can ask ApolloSRM to run a report, surface chronic absentees, or draft outreach — in one sentence, governed by the calling user's permissions. The same tools the AI crew uses are the tools you can build on. That is the unfair advantage: almost no career-school platform can say it.