Home

Welcome to the sas-score-mcp-serverjs wiki!

SAS-SCORE-MCP-SERVERJS MCP Server Tools Reference

Overview

The @sassoftware/sas-score-mcp-serverjs package provides a comprehensive set of tools for interacting with SAS Viya environments through the Model Context Protocol (MCP). This document catalogs all 24+ available tools organized by functional category.

All tools are registered as sas-score-

Start MCP Server

If using stdio transport, most of the mcp clients will start the server automatically. But for http transport, the mcp server must be started.

If running locally

npx @sassoftware/sas-score-mcp-serverjs@latest

The mcp is also available as a docker image. Add or remove the env variables as needed.

docker run -p 8080:8080 --name sasscore -e VIYA_SERVER=<yourviyaserver> -e AUTHFLOW=oauth ghcr.io/sassoftware/sas-score-mcp-serverjs:latest

Configuration

Typically these are set either in the .env file or as environment variables or as command line options(if using npx). You will need only a subset of these for the different [transport,authentication] schemes

Required Options

VIYA_SERVER=url for Viya server

Options with defaults

These can be customized

AUTHFLOW=oauth|oauthclient|bearer|sascli|token|password
- Authentication method. Default is oauth
CLIENTID=vscodemcp
- Clientid for oauth and oauthclient AUTHFlOW. Must be PKCE clientid.
MCPTYPE=http|stdio
- The transport protocol for the mcp server.
MCPHOST=http://localhost:8080
- URL of the mcp server. If using remote mcp server, set this to remote MCP server
PROFILE=~/.sas
- profile name used by sas-cli to store the tokens
PORT=8080
- set it to what fits your environment.
CASSERVER=cas-shared-default
- Set to a valid cas server
COMPUTECONTEXT="SAS Job Execution compute context"
- Use one that is appropriate

Clientid specifications

If using remote mcp server, change the url in redirect to the remote url

{
  client_id: 'vscodemcp',
  scope: [ 'openid' ],
  resource_ids: [ 'none' ],
  autoapprove: true,
  authorized_grant_types: [ 'authorization_code' ],
  access_token_validity: 86400,
  allowpublic: true,
  redirect_uri: [ 'http://localhost:8080/callback' ]
}

OauthClient Flow. Clientid with redirect appropriate for the client. Some examples are shown below. Note that the explicit port used by github copilot is not guaranteed.

github copilot: http://127.0.0.1:33418/
claude: https://claude.ai/api/mcp/auth_callback,https://claude.ai/api/auth/callback

Agent and Skills

This agent provides a unified interface for interacting with SAS Viya resources, including models, jobs, tables, and scoring endpoints. It automates the process of verifying, executing, and formatting requests, ensuring robust and consistent workflows for analytics and machine learning operations.

Operating Model

Every Request Follows Three Steps

VERIFY — Use find-resources to verify target resources exist
EXECUTE — Use the appropriate execution tool (sas-score-read-table, sas-score-mas-score, sas-score-run-jobdef, sas-score-scr-score, etc.)
FORMAT — Merge results and return to user

Request Classification

When you receive a SAS request, classify it using request-routing skill:

Type	Trigger	Strategy	Tool
Find	"find", "locate", "exists"	find-resources	sas-score-find-library, sas-score-find-table, etc.
Read	"read", "show", "fetch", "query", "how many", "count by"	read-strategy	sas-score-read-table, sas-score-sas-query
Score	"score", "predict", "run model"	score-strategy	sas-score-mas-score, sas-score-run-jobdef, sas-score-scr-score
List	"list", "show all"	—	sas-score-list-libraries, sas-score-list-tables, sas-score-list-models, etc.

Key Principles

1. Always Verify Before Executing

Exception: SCR models can score without pre-verification.

Verify resources exist (find-resources)
  ↓
Execute action (sas-score-read-table, sas-score-mas-score, sas-score-run-jobdef, sas-score-scr-score, etc.)
  ↓
Merge and format results

2. Determine Server for Tables

Every table operation must explicitly determine whether the table is in CAS or SAS:

CAS tables: Caslib.table (Casuser, Public, Samples, Formats, etc.)
SAS tables: LIBREF.table (SASHELP, WORK, SASUSER, etc.)

Use find-resources to determine server if not specified by user.

3. Explicit Model Type

If model type is ambiguous, default to MAS:

score with model X → MAS (default)
score with model X.mas → MAS
score with model X.job → Job
score with model X.jobdef → JobDef
score with model X.scr → SCR (no pre-verification)

4. No Invention

Never invent resource names, identifiers, servers, or model types. Always verify or ask.

Ambiguous Terms

These terms are overloaded in SAS and must be clarified:

model: MAS, Job, JobDef, or SCR?
score/scoring: Running a model on data (not code coverage)
job: SAS Job or SAS JobDef?
table: CAS table or SAS dataset? Which library?
resource: Library, table, model, job, or jobdef?
read/query: Raw row read or aggregation?

When ambiguous, ask one focused clarifying question.

Strategy Reference

See the strategies folder:

request-routing — Universal three-step workflow + examples
find-resources — How to verify resources exist
read-strategy — How to read/query tables
score-strategy — How to score/predict
sas-score-mcp-serverjs-agent — Main orchestration agent

Response Format

After completing a SAS Viya request, append a brief Strategy Summary:

---

**Strategy Summary:**
- **Classification**: [Request type you identified]
- **Verification**: [Resources verified or verification skipped]
- **Tool Used**: [Primary tool(s) invoked]
- **Routing Decision**: [Why you chose this path]

Example:

**Strategy Summary:**
- **Classification**: Score request with inline scenario
- **Verification**: Found job simplejob
- **Tool Used**: sas-score-run-jobdef
- **Routing Decision**: Job type (.job suffix) → run-jobdef tool

Error Handling

Error	Action
Resource not found	Ask user to verify name/spelling and server
Server ambiguous	Use find-resources to determine CAS vs SAS
Model type unclear	Ask: "Is this a MAS, Job, JobDef, or SCR model?"
Table column mismatch	Ask user for column → input mapping
Empty result	Ask user to adjust filter or criteria

Examples

Example 1: Score Inline Scenario

User: "score a=1, b=2 with model simplejob.job"

Process:

Classify: Score request, inline scenario, job type
Verify: Find job simplejob → Found ✓
Execute: sas-score-run-job({ name: "simplejob", scenario: { a: 1, b: 2 } })
Result: { a: 1, b: 2, c: 3 }

Example 2: Query with Aggregation

User: "average MSRP by make for cars from USA in sashelp.cars"

Process:

Classify: Read request, analytical query
Verify: SASHELP is default SAS library ✓
Execute: sas-score-sas-query({ table: "SASHELP.cars", query: "average MSRP by make where origin='USA'", ... })
Result: Aggregated data by make

Example 3: Read Rows

User: "read first 20 customers from Public"

Process:

Classify: Read request, raw row read
Verify: Find table customers in Public → CAS ✓
Execute: sas-score-read-table({ lib: "Public", table: "customers", server: "cas", limit: 20 })
Result: 20 customer rows

Example 4: Score Table Rows

User: "score all active customers with model risk_model"

Process:

Classify: Score request, table rows
Verify: Find model risk_model → MAS ✓, Find table customers → CAS ✓
Execute:
- Read: sas-score-read-table({ where: "status='active'" })
- Score: sas-score-mas-score({ model: "risk_model", scenario: {each row} })
Result: Rows with risk_score appended

Quick Reference: Tools

Category	Tool	Purpose
Find	sas-score-find-library	Verify library exists
	sas-score-find-table	Verify table exists + determine server
	sas-score-find-model	Verify MAS model exists
	sas-score-find-job	Verify job exists
	sas-score-find-jobdef	Verify jobdef exists
Read	sas-score-read-table	Get raw rows from table
	sas-score-sas-query	Query with aggregations
Score	sas-score-mas-score	Score with MAS model
	sas-score-run-jobdef	Run job or jobdef
	sas-score-scr-score	Score with SCR model
List	sas-score-list-libraries	Browse all libraries
	sas-score-list-tables	Browse tables in library
	sas-score-list-models	Browse MAS models
	sas-score-list-jobs	Browse jobs
	sas-score-list-jobdefs	Browse jobdefs

When to Ask for Clarification

Ask one focused question if:

Resource name is missing or ambiguous
Model type is not specified (.mas/.job/.jobdef/.scr)
Table library is not specified
Request involves "model" without type context
User is asking for aggregation but column mapping is unclear
Scoring inputs don't match model signature

Do NOT guess. Ambiguity is better resolved than guessed.

When to Verify

Always verify before executing, except:

List operations (list-* tools don't need pre-verification)
SCR scoring (SCR can score without pre-check)

Summary

This is a three-step, reliable system:

Classify the request (find/read/score/list)
Verify resources exist (except list and SCR)
Execute with the right tool
Format results + append Strategy Summary

No ambiguity. No guessing. No invented resource names.

Model Management & Scoring

list-models

Enumerate models published to MAS (Model Aggregation Service).

Parameters:

limit (number, default: 10): Number of models to return
start (number, default: 1): 1-based offset for pagination

Usage:

"list models"
"show models"
"list 25 models"
"next models" (pagination)

Example:

list models
list 25 models

find-model

Locate a specific model deployed to MAS.

Parameters:

name (string, required): Exact model name

Usage:

"find model churnRisk"
"does model creditScore exist"
"is model sales_forecast deployed"

Example:

find model myModel

model-info

Retrieve detailed metadata for a deployed model including input/output variables, data types, and constraints.

Parameters:

model (string, required): Model name as published to MAS

Returns:

Input variable metadata (names, types, roles, ranges)
Output variable information
Model type and description

Usage:

"What inputs does model X need?"
"Describe model myModel"
"Show the variables for sales_forecast"

Example:

model-info model=churnRisk

model-score

Score user-supplied scenario data using a MAS-published model.

Parameters:

model (string, required): Model name
scenario (string | object | array, required): Data to score
uflag (boolean, optional): Prefix model fields with underscore

Scenario formats:

Comma-separated: "x=1, y=2"
Object: {x: 1, y: 2}
Array: [{x: 1, y: 2}, {x: 3, y: 4}]

Usage:

"Score this customer with model churnRisk"
"Run model creditScore with age=45, income=60000"

Example:

model-score model=mycoolmodel scenario={x:1,y:2}
model-score model=cancer1 scenario="age=45, sex=M, tumor=stage2"

scr-info

Return input/output schema and metadata for an SCR (Score Code Runtime) model.

Parameters:

name (string, required): SCR model identifier (URL or name)

Returns:

Input variables (names, types, required/optional)
Output variables (predictions, probabilities, scores)

Example:

scr-info name="https://scr-host/models/loan"

scr-score

Score a scenario using an SCR container model.

Parameters:

url (string, required): SCR model identifier (URL)
scenario (string | object | array, optional): Input values

Usage:

Run scrInfo first to inspect expected inputs
Omit scenario to get model metadata

Example:

scr-score url="loan" scenario="age=45, income=60000"
scr-score url="https://scr-host/models/loan" scenario={age:45, income:60000}

Library Management

list-libraries

Enumerate CAS or SAS libraries.

Parameters:

server (cas|sas, default: 'cas'): Target environment
limit (number, default: 10): Page size
start (number, default: 1): 1-based offset
where (string, optional): Filter expression

Usage:

"list libs"
"list libraries"
"show cas libs"
"list sas libs"

Example:

list libraries
list sas libs
show me 25 cas libraries

find-library

Locate a specific CAS or SAS library.

Parameters:

name (string, required): Exact library name
server (cas|sas, default: 'cas'): Target environment

Usage:

"find library Public"
"does library SASHELP exist"
"is PUBLIC library available in cas"

Example:

find lib Public
find library sasuser in sas

Table Operations

list-tables

Enumerate tables within a specific CAS or SAS library.

Parameters:

lib (string, required): Library to inspect
server (cas|sas, default: 'cas'): Target environment
limit (number, default: 10): Page size
start (number, default: 1): 1-based offset

Usage:

"list tables in Samples"
"show tables in sashelp"
"list 25 tables in Public"

Example:

list tables in samples
show 25 tables in sashelp

find-table

Locate a table in a specified library.

Parameters:

lib (string, required): Library to search in
name (string, required): Table name or substring
server (cas|sas, default: 'cas'): Target environment

Usage:

"find table iris in Public library in cas"
"find table cars in sashelp in sas server"

Example:

find table iris in Public
find table cars in sashelp in sas

table-info

Return metadata about a table including columns, types, and statistics.

Parameters:

table (string, required): Table name
lib (string, required): Library containing the table
server (cas|sas, default: 'cas'): Target environment

Returns:

Column metadata (name, type, label, formats)
Table statistics (row count, file size, timestamps)

Usage:

"describe table cars in Public"
"info on table mydata in mylib"

Example:

table-info table=cars lib=Public
describe table air in lib sashelp on sas server

read-table

Retrieve rows from a table in a CAS or SAS library.

Parameters:

table (string, required): Table name
lib (string, required): Library containing the table
server (cas|sas, default: 'cas'): Target environment
start (number, default: 1): Starting row (1-based)
limit (number, default: 10): Maximum rows to return
where (string, optional): SQL-style WHERE clause
format (boolean, default: true): Return formatted or raw values
row (number, optional): Read a specific row

Usage:

"read table customers"
"show me 10 rows from sales"
"read from orders where status = 'shipped'"

Example:

read table cars in Samples
show 25 rows from customers
read orders where status = 'shipped' limit 50
read row 15 from employees in mylib on sas

sas-query

Execute SQL queries on SAS tables using PROC SQL.

Parameters:

table (string, required): Table in format libname.tablename
query (string, required): Natural language query
sql (string, optional): Generated SQL SELECT statement
job (string, default: 'sas_sql_tool'): Job to run query on

Workflow:

User provides natural language query
Convert to SAS PROC SQL SELECT statement
Execute and return results

Example:

sasquery table=mylib.clm_dental query="Total paid amount, unique patients by procedure code"
sasquery table=mylib.students query="How many students in each year as percentage"

Job Management

list-jobs

Enumerate SAS Viya job assets.

Parameters:

limit (number, default: 10): Number of jobs to return
start (number, default: 1): 1-based offset
where (string, optional): Filter expression

Usage:

"list jobs"
"show jobs"
"list 25 jobs"

Example:

list jobs
list 25 jobs

find-job

Locate a specific SAS Viya job.

Parameters:

name (string, required): Exact job name

Usage:

"find job cars_job_v4"
"does job sales_summary exist"
"verify job ETL_Daily"

Example:

find job cars_job_v4

run-job

Execute a job on a SAS Viya server.

Parameters:

name (string, required): Job name
scenario (string | object, optional): Input parameters

Returns:

Log, listing, and tables (depending on job definition)

Example:

run job xyz param1=10,param2=val2
run-job myjob scenario a=10,b=20
job myjob scenario a=10,b=20

list-jobdefs

Enumerate SAS Viya job definition assets.

Parameters:

limit (number, default: 10): Number to return
start (number, default: 1): 1-based offset
where (string, optional): Filter expression

Usage:

"list jobdefs"
"show job definitions"

Example:

list jobdefs
list 25 jobdefs

find-jobdef

Locate a specific job definition.

Parameters:

name (string, required): Exact jobdef name

Usage:

"find jobdef cars_job_v4"
"does jobdef ETL exist"

Example:

find jobdef metricsRefresh

run-jobdef

Execute a job definition on a SAS Viya server.

Parameters:

name (string, required): Job definition name
scenario (string | object, optional): Input parameters

Returns:

Log, listing, and tables

Example:

run-jobdef xyz param1=10,param2=val2
jobdef myjobdef scenario a=10,b=20

Program Execution

run-program

Execute arbitrary SAS code or stored programs on a SAS Viya server.

Parameters:

src (string, required): SAS code or .sas filename
folder (string, optional): Server folder path if src is a filename
scenario (string | object, optional): Input parameters
output (string, optional): Name of output table returned as JSON
limit (number, default: 100): Max rows from output table

Usage:

Direct code execution
Running stored .sas files
With input parameters and output capture

Example:

run program "data a; x=1; run;"
program "data work.a; x=1; run;" output=a limit=50
run program sample folder=/Public/models output=A limit=50
program sample folder=/Public/models scenario="name='John', age=45" output=a

run-macro

Submit and execute a SAS macro on a SAS Viya server.

Parameters:

macro (string, required): Macro name (without leading %)
scenario (string, optional): Parameters or SAS setup code

Scenario formats:

Comma-separated: "x=1, y=abc" → converted to %let statements
Raw SAS: "%let x=1; %let y=abc;" → passed through unchanged

Example:

run macro abc with scenario x=1, y=2
run macro summarize with scenario %let x=1; %let y=2;

Context & Configuration

set-context

Set the CAS and SAS server contexts for subsequent tool calls.

Parameters:

cas (string, optional): CAS server name
sas (string, optional): SAS compute context name

Returns:

Current CAS and SAS context values

Usage:

Switch between server environments
Check current context (call with no parameters)

Example:

set-context cas=finance-cas-server
set-context sas="SAS Studio Compute Context"
set-context (returns current context)

Utility Tools

Use this to verify that the mcp server is up and running.

deva-score

Compute a numeric score based on two input values using the formula: (a + b) × 42

Parameters:

a (number, required): First numeric input
b (number, required): Second numeric input

Returns:

Numeric result: (a + b) × 42

Usage:

"Calculate deva score for 5 and 10"
For sequences: chain calls left-to-right

Example:

deva-score a=5 b=10  // returns 630
deva-score a=1 b=2   // returns 126

Tool Categories Summary

Category	Tool Count	Tools
Model Management	6	list-models, find-model, model-info, model-score, scr-info, scr-score
Library Management	2	list-libraries, find-library
Table Operations	5	list-tables, find-table, table-info, read-table, sas-query
Job Management	6	list-jobs, find-job, job, list-jobdefs, find-jobdef, job-def
Program Execution	2	run-program, run-macro
Context & Config	1	set-context
Utilities	1	deva-score
Total	24

Common Patterns

Discovery → Inspection → Action

List tools to discover available resources
Find tools to locate specific items
Info tools to inspect metadata
Execution tools to perform actions

Pagination

Many list tools support pagination:

First page: { start: 1, limit: 10 }
Next page: { start: 11, limit: 10 }

Server Targeting

Tools that interact with data support server parameter:

'cas' - CAS server (default)
'sas' - SAS compute server

Scenario Input Formats

Tools accepting scenarios support multiple formats:

String: "x=1, y=2"
Object: {x: 1, y: 2}
Array: [{x: 1, y: 2}]

Notes

All tools are designed to work with SAS Viya environments
Authentication and connectivity are handled by the MCP server configuration
Tools are stateless unless context is explicitly set using set-context
Error handling returns structured error objects from the backend
Case sensitivity varies by backend (library/table names may be case-insensitive)

Document generated for @sassoftware/mcp-serverjs
Last updated: December 2024

Uh oh!

Home

SAS-SCORE-MCP-SERVERJS MCP Server Tools Reference

Overview

Table of Contents

Start MCP Server

Configuration

Required Options

Options with defaults

Clientid specifications

Agent and Skills

Operating Model

Every Request Follows Three Steps

Request Classification

Key Principles

1. Always Verify Before Executing

2. Determine Server for Tables

3. Explicit Model Type

4. No Invention

Ambiguous Terms

Strategy Reference

Response Format

Error Handling

Examples

Example 1: Score Inline Scenario

Example 2: Query with Aggregation

Example 3: Read Rows

Example 4: Score Table Rows

Quick Reference: Tools

When to Ask for Clarification

When to Verify

Summary

Model Management & Scoring

list-models

find-model

model-info

model-score

scr-info

scr-score

Library Management

list-libraries

find-library

Table Operations

list-tables

find-table

table-info

read-table

sas-query

Job Management

list-jobs

find-job

run-job

list-jobdefs

find-jobdef

run-jobdef

Program Execution

run-program

run-macro

Context & Configuration

set-context

Utility Tools

deva-score

Tool Categories Summary

Common Patterns

Discovery → Inspection → Action

Pagination

Server Targeting

Scenario Input Formats

Notes

Clone this wiki locally