Gecko AI Platform

The model is a setting, not a marriage.

Run agents on the AI providers you choose — with your own keys or ours — and make your SIS, CRM and proprietary systems agent-ready through hosted connectors and the open Model Context Protocol.

Map your estate with us Browse the catalogue
7model provider families
22+built-in integrations
45Gecko Engage MCP tools
Anysystem, via MCP
Vendor-neutral AI

Seven provider families. One governed catalogue.

Connect a provider and the platform asks it which models it can actually serve — including your own Azure deployments and Bedrock custom models. IT then enables exactly the models each tenant may use, and builders pick only from that list.

  • Bring your own keys so inference runs under agreements you already hold — or use Gecko-managed access for a faster start
  • Swap the model behind a live agent from a dropdown, without rebuilding it
  • Tenant defaults for chat, embedding and reranking models; agents inherit unless overridden
  • Reasoning effort and verbosity set per agent and per AI step, validated against the chosen model
Supported model providers and what each can serve
ProviderChatEmbeddingsNotes
OpenAI✓✓Custom base URL, organisation and project
Anthropic✓—Native web search, web fetch and code execution tools
Amazon Bedrock✓✓Foundation and custom models; reranking
Google Gemini✓✓Live model discovery
Microsoft Foundry✓✓Entra ID or API key; discovers OpenAI, Claude and DeepSeek deployments
OpenRouter✓✓The OpenRouter catalogue
Custom✓✓Any OpenAI-compatible endpoint, including self-hosted — gated by an origin allowlist
Why neutrality matters

The best model changes every quarter. Your platform shouldn’t.

Institutional AI shouldn’t be a one-way door. Neutrality is a procurement, resilience and governance decision — not a technical nicety.

Procurement leverage

When the model is a dropdown, you negotiate with vendors from strength — or route through Microsoft and AWS agreements you already hold.

Future-proofing

Adopt a better or cheaper model by changing a setting. Prompts, knowledge, tools and workflows stay exactly where they are.

Data governance

Institution-owned keys keep inference inside contracts you already signed — for example, Microsoft Foundry within your Microsoft tenancy.

Specialist AI

The right model for each job — not just chat.

Beyond chat and embeddings, the platform supports dedicated models for the steps where general-purpose models are a blunt instrument.

  • Reranking with Cohere, Amazon Bedrock rerankers or a Cohere-compatible endpoint — sharper retrieval before the model reads anything
  • Evaluation with TypeSafe — calibrated probabilities for choice, score and yes/no questions, ready for confidence-based routing
How workflows use them
The catalogue

Built for the estate you run — and whatever comes next.

22+ built-in integrations today, each with an in-console setup guide — with more added all the time, and unlimited reach through MCP, webhooks and custom code.

AI models · 7

OAOpenAI
chat · embeddings
AAnthropic
chat · native tools
BRAmazon Bedrock
chat · embed · rerank
GGoogle Gemini
chat · embeddings
MFMicrosoft Foundry
Azure deployments
OROpenRouter
chat · embeddings
{ }Custom endpoint
OpenAI-compatible

Specialist AI · 2

CoCohere
reranking
TSTypeSafe
evaluation

Higher-ed & CRM · 3

EEllucian Ethos
hosted MCP · SIS
SFSalesforce
hosted MCP · OAuth
GEGecko Engage
hosted MCP · 45 tools

Files & knowledge · 5

SPSharePoint
crawl · sync · files
ODOneDrive
sync · files
GDGoogle Drive
sync · files
DBDropbox
sync · files
S3Amazon S3
incl. S3-compatible

Messaging, voice & compute · 5

Microsoft Teams
agents · approvals
Slack
agents · approvals
TwTwilio
voice · SMS
ReResend
email
▲Vercel Sandbox
custom code
And many more

Don’t see your system? Connect it anyway.

New built-in integrations ship regularly — and the catalogue is never the limit. Anything with an API can be made agent-ready, with the same per-agent grants, guardrails and tracing.

Ask about a system How MCP works
  • Any MCP serverUse an existing MCP server or build your own for proprietary systems — timetabling, accommodation, library, finance.
  • Webhooks & APICall any approved endpoint from a workflow, or trigger Gecko from your own systems.
  • Custom CodeWrite the connector logic in a sandboxed step, with secrets bound at run time.

Product names are trademarks of their respective owners and are used to identify supported integrations.

Institutional systems

Your SIS and CRM, agent-ready — safely.

Gecko hosts production MCP servers for the systems universities run on, with security models built for an IT architect’s scrutiny.

Ellucian Ethos

Read and query Ethos resources — the integration layer in front of Banner and Colleague — through Ethos’s governed Proxy API, never SQL.

  • Regional endpoints for the US, Canada, Ireland and Australia
  • Every request carries a short-lived, Gecko-signed ES256 token bound to the tenant and scope
  • When Gecko manages the connection, your Ethos API key is never forwarded to the MCP server

Salesforce

Connect production and sandbox orgs with OAuth, and let the hosted Salesforce MCP server act with each person’s own Salesforce permissions through per-user sign-in.

  • OAuth 2.0 with PKCE
  • Per-user sign-in — agents never exceed the user’s access
  • Tools granted to agents individually

Gecko Engage

The whole student-engagement suite becomes agent-callable: 45 tools across contacts, calls, events, forms, campaigns, templates, users and files — plus read-only database queries.

  • Find and create contacts, log activity, send email and SMS
  • Book contacts onto event sessions and submit forms
  • Region-aware authentication
Model Context Protocol

Your 15-year-old timetabling system, agent-ready.

MCP is the open standard for exposing tools to AI agents. Register any remote MCP server — yours, a vendor’s or one Gecko builds with you — and IT decides exactly which agents may use which tools.

  • Any remote server over Streamable HTTP or SSE
  • Authentication with headers, a saved integration, or per-user OAuth sign-in with PKCE and dynamic client registration
  • Per-agent, per-tool grants — nothing is granted by default
  • Test bench — inspect each tool’s schema, fill a generated form and run a live call before any agent touches it
  • Portable investment — tools you build work with any MCP client, not just Gecko

Your SIS, your CRM and the system nobody else integrates with — all agent-ready through one open standard. Agents get tools one checkbox at a time.

Setup & credentials

Connected with in-console guides, not consulting hours.

Every provider ships with a numbered, deep-linked setup walkthrough — and credentials are handled the way your security team expects.

Guided setup

Step-by-step guides for every integration, with generated webhook and OAuth callback URLs to paste where they belong.

Generated Slack manifest

The console writes the complete Slack app manifest for you — paste it into Slack and go. Teams uses a guided Azure Bot registration.

Write-only credentials

Keys and secrets are AES-256-GCM encrypted at rest and never returned after save — not to the UI, not to the API, not to admins.

PlatformKnowledge & retrieval BuildDevelopers TrustSecurity & governance
Map your estate

Tell us what you run. We’ll show you how it connects.

Bring your model agreements, SIS, CRM and the systems nobody else integrates with. We’ll map the fastest governed path to each.

  • Model strategy: your keys, ours, or both
  • SIS and CRM connection options
  • An MCP plan for proprietary systems
Book a conversation