SAS Viya MCP Server

A Model Context Protocol (MCP) server for executing SAS code, training AutoML projects, scoring models and so much more for SAS Viya environments.

Features

• 91 tools across 10 selectable tiers, spanning the Analytics Life Cycle on SAS Viya
• Prompt Templates for improving your SAS Code
• OAuth2 authentication with PKCE flow
• HTTP-based MCP server compatible with MCP clients

Articles & Videos

Here you can find getting articles on how to use and integrate the SAS MCP Server in different tools and what to build with it:

• From REST APIs to AI Agents: Why the SAS Viya MCP Server Matters
• Connecting GitHub Copilot to SAS Viya with the SAS Viya MCP Server
• Bring Your Own Key: SAS Viya MCP Server with GitHub Copilot CLI
• Putting the SAS Viya MCP Server to Work in GitHub Copilot
• Connecting Claude Code CLI to SAS Viya with the SAS Viya MCP Server
• Putting the SAS Viya MCP Server to Work in Claude Code CLI
• Integration with SAS Retrieval Agent Manager (RAM)

Getting Started

Prerequisites

• Required

  • Python 3.12+
  • uv 0.8+
  • SAS Viya environment with compute service
  • Setup the Viya environment for MCP
    • See configuration.md

• Optional

  • Docker: refer to container setup
  • Kubernetes: sample manifests (Contour or nginx) and a Helm chart in deploy/

Installation

1. Clone the repository:

git clone <repository-url>
cd sas-mcp-server

2. Install dependencies

uv sync

NOTE: This will by default create a virtual environment called .venv in the project's root directory.

If for some reason the virtual environment is not created, please run uv venv and then re-run uv sync.

Usage

1. Configure environment variables:

cp .env.sample .env

Edit .env and set

VIYA_ENDPOINT=https://your-viya-server.com

2. Start the MCP server (see Choosing a deployment mode below):

Option A: HTTP mode (pre-run the server, connect from MCP client)

uv run app

The server will be available at http://localhost:8134/mcp by default. Authentication is handled via OAuth2 PKCE flow in the browser.

Option B: Stdio mode (MCP client starts the server on demand)

Authenticate once. Two equivalent options:

# Option B1 — if you have the SAS Viya CLI installed:
sas-viya auth loginCode

# Option B2 — built-in helper, no external CLI needed (Viya 2022.11+):
uv run sas-mcp-login

Both flows write an access token to a local cache (~/.sas/credentials.json and ~/.sas-mcp-server/credentials.json respectively); the stdio server reads whichever it finds. When the token expires, re-run the same command.

Then configure your MCP client to launch the server directly (see below).

Option C: Docker / Podman (containerized deployment)

Pull the pre-built image from GitHub Container Registry:

docker pull ghcr.io/sassoftware/sas-mcp-server:latest
docker run -e VIYA_ENDPOINT=https://your-viya-server.com -p 8134:8134 ghcr.io/sassoftware/sas-mcp-server:latest

Or build locally from source:

docker build -t sas-mcp-server .
docker run -e VIYA_ENDPOINT=https://your-viya-server.com -p 8134:8134 sas-mcp-server

Available image tags:

• latest — most recent tagged release
• <major>.<minor>.<patch> (e.g. 1.0.0) — specific release
• <major>.<minor> (e.g. 1.0) — latest patch of a minor release
• edge — tip of main (unreleased, for testing)
• sha-<short> — pinned to a specific commit

Programmatic clients with a pre-existing Viya token

If your caller already holds a Viya access token (e.g. an automation script that obtained one via the SAS Viya CLI), start the HTTP-mode server with ALLOW_RAW_BEARER=true and pass the token directly:

curl -H "Authorization: Bearer $VIYA_TOKEN" http://localhost:8134/mcp ...

The server validates the token against Viya's JWKS and uses it upstream as-is, bypassing the MCP JWT swap. The default OAuth2 PKCE flow keeps working alongside — both client types share the same /mcp endpoint.

If your Viya APIs are intentionally exposed without auth (for example, a local/dev Compute API endpoint), set VIYA_AUTH=false to bypass all SASLogon/OAuth flows in both HTTP and stdio modes. In this mode the server sends upstream requests without an Authorization header.

If your compute deployment does not expose /compute/contexts and only supports a fixed session, set COMPUTE_SESSION_ID=<session_id>. The compute tools will use that session directly instead of creating context-backed sessions.

Choosing a deployment mode

Quick guidance:

• Starting out or exploring? Use stdio — one sas-viya auth loginCode or uv run sas-mcp-login, then your MCP client manages the server lifecycle.
• Need secure, interactive auth? Use HTTP — no stored passwords, each user authenticates via browser.
• Deploying for a team or on a server? Use Docker — portable, no Python dependency on the host, easy to integrate with orchestrators.
• Running it for a whole organisation? Use Kubernetes — sample manifests and a Helm chart are in deploy/, including the routing the OAuth flow needs for either Contour (the chart's default, and the only one that can mount the server under a path prefix on an existing hostname) or ingress-nginx.
• Using Gemini CLI? Use stdio — Gemini CLI does not support HTTP mode or browser-based OAuth. See Gemini CLI configuration.
• Installing from a client's server catalogue? That path runs the published container in stdio mode (app-stdio), not as an HTTP server, so it authenticates from your ~/.sas token cache — which has to be mounted into the container at /app/.sas.

Limiting exposed tools (tiers)

Tools are grouped into numbered tiers. By default the server exposes all of them; set MCP_TIERS to expose only a subset — handy for keeping a client's tool list small and focused, or hiding capabilities a deployment shouldn't offer. Accepts ranges and comma lists (e.g. MCP_TIERS=0-4 or MCP_TIERS=0,1,6,7); unset means all tiers.

# Example: expose only compute/discovery/data-ops and reporting
MCP_TIERS=0-3 uv run app

Read-only mode

Set MCP_READ_ONLY=true to expose only tools that neither change server-side state nor cause server-side work — 50 of the 91 tools. Withheld tools are never registered, so they are absent from the client's tool list entirely: the model cannot see them, so it cannot attempt them.

This is a filter over the tiers, not a tier of its own — the read/write split cuts across every tier (Tier 3 has both get_report and delete_report). The two settings compose:

# Every read tool, all tiers
MCP_READ_ONLY=true uv run app

# Read tools of the reporting and decisioning tiers only
MCP_TIERS=3,7 MCP_READ_ONLY=true uv run app

The definition is strict: a tool qualifies only if it can neither write nor start work. Beyond the obvious create/update/delete tools, that withholds: