Evaluating Flexday AI
Fit with your technology stack
An inventory of what Flexday AI connects to - identity providers, Teams, ServiceNow, email, clouds, storage, vector stores, AI models, APIs and data import.
Written for
- Technical
Last reviewed
An enterprise architect's first question about a new platform is what it will need from the estate you already run, and what it can connect to. This page is the inventory: each category says what is supported today, how it is configured, and which page explains it. It lists only what the platform's code implements; anything not listed is not supported in this release.
At a glance
- Your identity, not new accounts. Builders sign in through Microsoft Entra ID, Amazon Cognito or any OpenID Connect or SAML provider; the people who use what is built can sign in through any of 30 provider presets or nine credential mechanisms.
- Two clouds. Drivers for AWS and Azure services, from object storage to secret vaults.
- Models from Anthropic and Azure AI Foundry. Anthropic Claude and Azure AI Foundry GPT models, with embeddings from Voyage AI, OpenAI or Azure AI Foundry.
- Open interfaces. A documented API with an OpenAPI description, per-Solution gateway endpoints, and signed webhooks.
The inventory
| Category | What is supported | How it is configured | Read more |
|---|---|---|---|
| Studio sign-in | Microsoft Entra ID (workforce), Microsoft Entra External ID, Amazon Cognito, any OpenID Connect provider, and SAML through a user pool | A workspace's sign-in setting: pooled, a dedicated user pool, or your own provider | Identity and access |
| Sign-in for apps and APIs you build | 30 provider presets: 29 named providers (among them Entra ID, Okta, Auth0, Ping, AD FS, Cognito, Keycloak and Google) plus any OpenID Connect provider; nine mechanisms in all, including SAML federation, API keys, signed requests, Basic, token introspection and client certificates | An Identity attached to a Solution's gateway | Identity |
| Microsoft Teams | A published Agent as a Teams Bot, in one-to-one chats, group chats and team channels; messages the Bot starts itself reach only people and chats that have already messaged or installed it, and stop for anyone who sends STOP in a one-to-one chat | A Bot, registered by the platform or with your own Azure bot registration | Bot |
| ServiceNow | Nine actions: create, update, comment on, assign, resolve, get and search tickets; find users and groups | An application Connection with OAuth client credentials or a user and password, used by a Flow's Integration step | Integration |
| Other business applications | Any HTTP API | A Flow's HTTP request step or an Agent's HTTP grant, with a Connection for its credential | Connection |
| The platform's own mail sender, or your own SMTP server | A deployment setting, or an SMTP Connection in the Solution | Workflow automation | |
| Cloud: AWS | The reference deployment: ECS Fargate, Aurora PostgreSQL, ElastiCache, S3, EFS, Cognito, SES, Secrets Manager and KMS | Terraform for the whole environment | Reference deployment on AWS |
| Cloud: Azure | The reference deployment: Container Apps, Front Door, PostgreSQL Flexible Server, Managed Redis, Blob Storage, Azure Files, Key Vault and Entra ID, with drivers for Blob Storage, Key Vault, Entra ID and Azure AI Foundry | Terraform for a dedicated deployment in your subscription (one development environment so far) | Reference deployment on Azure |
| Object storage | Amazon S3, Azure Blob Storage, or a local file system for a single server | One storage setting per deployment | Data architecture |
| Vector store for document search | PostgreSQL with pgvector (the default), or Qdrant | A deployment default, fixed per Doc Base when it is created; text and the keyword index always stay in PostgreSQL | Doc Base |
| Secret storage | AWS Secrets Manager, Azure Key Vault, or Flexday AI's own encrypted storage | Set by the deployment; a workspace does not choose it | Secrets and encryption |
| AI models | Anthropic Claude (Opus, Sonnet and Haiku models); GPT models through Azure AI Foundry; embeddings from Voyage AI, OpenAI or Azure AI Foundry | A model per use case, with workspace and per-resource overrides | AI models |
| APIs | The platform's own REST API, described by a generated OpenAPI document; per-Solution Flex Gateway endpoints that run a saved query, a Flow, an Agent or a file operation | Endpoints are created in the Studio, each with its own sign-in rule | Flex Gateway |
| Webhooks | Inbound: a Flow started by a signed request to its own unguessable address. Outbound: any Flow step that calls an HTTP API | A webhook trigger on the Flow | Flow |
| Data import | Sample data as JSON rows or a CSV attached in the Builder; a database schema as SQL DDL or a definition document; a spreadsheet (.xlsx or .ods) mapped onto an existing Fact Base through the API (the Studio has no screen for this); a Fact Base export bundle | The Builder, a Fact Base's data load, or an import | Fact Base |
| Documents | PDF, Word (.docx), CSV, TSV, plain text, Markdown, HTML, JSON and JSON Lines, or a crawl of a website, up to 20 pages, two links deep | Uploaded to, or crawled into, a Doc Base | Doc Base |
Not supported in this release
- Messaging channels other than Microsoft Teams. Slack, WhatsApp, Telegram and SMS are names the platform reserves, and each is refused as not available.
- Business-application integrations other than ServiceNow through the Integration step. Other systems are reached through HTTP steps and Connections.
- Older binary Office formats, such as .xls and .doc, and Excel workbooks as Doc Base documents.
- Bringing your own encryption key for the platform's database or object storage, or pointing the shared service at your own vault for Secret Variables. A dedicated deployment runs in your own cloud account and can be set to seal Secret Variables in a vault there.