Open to your client’s stack.

The solution connects to what the company already uses: data platforms, identity, tools and agents. We use open standards and build adapters and connectors where the project requires. Governance already comes with the product.

Talk about my stackSee how it connects

Executive view

What changes in practice for decision-makers when the solution is open to the stack the company already has.

Your stack is the starting point
The data stays where it is. The fabric connects to the platforms the company already runs, with no migration required.
No lock-in
Connections use open standards (MCP, OpenID Connect, SQL, HTTP). Replacing a piece of the stack does not force a rebuild of the solution.
Each person with their own account
Sign-in uses the company’s identity provider. The person uses the tools and agents, with the access they already have.
Scope and timeline per project
Each integration has requirements, owners and test criteria defined beforehand. We say clearly what is available today and what depends on a project.

Your client’s stack, by category

The categories of a data and AI stack, what the solution connects today and what it connects per project.

Available in the productPer project

  1. Hyperscalers

    The cloud the stack already runs on.

    • AWSPer project
    • Microsoft AzurePer project
    • Google Cloud (GCP)Per project

    How we connect. Network, identity and costs are the client’s. The connection is defined in each project.

  2. Data platforms

    Data warehouse and lakehouse where the data lives.

    • SnowflakeAvailable in the product
    • Databricks (Unity Catalog)Available in the product
    • PostgreSQLAvailable in the product
    • Microsoft FabricPer project
    • Google BigQueryPer project
    • Amazon RedshiftPer project

    How we connect. Data adapter with a descriptor and a driver. Metadata reads discover the scope before any use.

    Arcogi is a Registered partner in the Snowflake Partner Network.

  3. Integration and transformation

    Ingestion, transformation and orchestration.

    • dbtPer project
    • FivetranPer project
    • Apache AirflowPer project
    • InformaticaPer project

    How we connect. The solution does not replace ingestion: it consumes what the stack already prepares. The connection is per project.

  4. Data governance and catalog

    Catalog, lineage and quality.

    • The product’s data catalog, domains and contractsAvailable in the product
    • Microsoft PurviewPer project
    • CollibraPer project
    • AlationPer project
    • AtlanPer project

    How we connect. Governance is already part of the product: data catalog, domains and contracts. With the client’s governance platform, it coexists and connects per project.

  5. BI and analytics

    Where people analyze and decide.

    • Power BIPer project
    • TableauPer project
    • LookerPer project
    • QlikPer project

    How we connect. Data products are consumed by the client’s analytics tools. We use MCP when the tool offers it.

  6. AI, models and agents

    Model providers, agents and clients.

    • Anthropic (Claude)Available in the product
    • OpenAIAvailable in the product
    • Third-party agents (A2A)Available in the product
    • MCP clients, such as Claude CodeAvailable in the product
    • Azure OpenAIPer project
    • Amazon BedrockPer project
    • Google Vertex AIPer project

    How we connect. Model providers configured per organization. Third-party agents come in over A2A and clients over MCP.

  7. Identity and security

    Who the person is and what they may access.

    • Corporate OpenID Connect providerAvailable in the product
    • Microsoft Entra IDPer project
    • OktaPer project

    How we connect. OpenID Connect with the corporate provider. Permissions come from the platform, not the token.

  8. Applications and service management

    Source systems and work management.

    • Jira (via MCP)Available in the product
    • ServiceNowPer project
    • SAPPer project
    • SalesforcePer project
    • OraclePer project
    • Microsoft DynamicsPer project

    How we connect. Reads through an adapter or an MCP server. The connection is defined in each project.

Product and company names belong to their owners and appear only to indicate the client’s stack that the solution can connect to. They do not indicate partnership, certification or endorsement, except the Registered partnership with Snowflake, stated under Data platforms.

Technical view

The protocols and standards the product uses at each layer.

LayerStandard or protocolHow we use it

Identity

OpenID Connect · Authorization Code with PKCE

The person signs in with the corporate provider’s account. The token is validated by signature (JWKS), issuer, audience and nonce. Permissions come from the platform, not from token fields.

Agent and tool access

MCP (JSON-RPC over HTTP, with SSE) · OAuth 2.1 with PKCE S256

An MCP server exposes agents and tools to clients such as Claude Code, with the initialize, tools/list and tools/call methods. The client is registered and the person authorizes on first connection.

External tools

MCP client for remote servers

We connect third-party MCP servers, such as Jira. Each tool is qualified before use and its definition is verified by hash: if it changes on the remote server, the tool is taken offline until reviewed again.

Third-party agents

A2A (message/send) · signed Agent Card (JWS, Ed25519)

The platform is both an A2A client and server: it delegates to vendor agents and receives partner calls, with the agent card published and signed.

Data

Adapters: Snowflake (SQL API), Databricks (Unity Catalog), PostgreSQL

Authentication by key pair (JWT), programmatic access token or service-principal token, depending on the platform. Secrets are kept in a vault. Metadata reads discover the scope.

API integration

Versioned HTTP API (/api/v1) · webhook

Commands and queries over the API. Agents can be triggered manually or by webhook.

This describes what the product implements today. Network, identity and authentication details are confirmed in each environment.

How an adapter or connector comes in

An adapter talks to a data platform. A connector links tools and agents over MCP or A2A. The path is the same for both.

  1. 01

    Stack discovery

    We map the client’s platforms, identity, networks and access rules.

  2. 02

    Descriptor

    Each adapter type is described as data: connection fields, authentication methods, required privileges and tests. A new type enters as a new description.

  3. 03

    Driver or connector

    A driver talks to the platform. When the platform offers MCP, we use the MCP connector.

  4. 04

    Test and scope

    Connection test, check of the minimum privileges and scope discovery before any use.

  5. 05

    Operation

    The connection goes into operation with an owner, usage tracking and periodic review.

Start with your stack.

Bring the design of your environment. We tell you what connects today, what needs a new adapter and the timeline.

Talk about my stackCookbooks

Available in the product and validated in a demonstration environment with synthetic data. AWS, GCP, Microsoft and the other platforms are connected per project, with agreed scope and test criteria. We do not declare an integration with a vendor before it exists.

Which part of your operation needs to move forward?

Start with a recurring decision or an outcome that matters. Together we identify the data, people and first capability required.

Talk about my stack