@squirex.dev/mcp-server
OfficialAutomates PR scanning with SARIF output via SquireX GitHub App, providing security checks on every pull request.
Scans Salesforce Agentforce metadata (Apex, Flows, GenAI, etc.) for 61+ security and operational risks, and can push test specs to Salesforce Testing Center.
Analyzes Slack bot configurations for missing DLP guards in Agentforce integrations.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@@squirex.dev/mcp-server run a full security scan on my agentforce metadata"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
@squirex.dev/mcp-server
SquireX MCP Server ā Agentforce Capability Scanner for AI Coding Agents
Model Context Protocol server exposing the SquireX Agentforce Capability Scanner to AI-powered developer tools. Works with Claude Code, Gemini, GitHub Copilot, and any MCP-compatible IDE.
Why SquireX?
Salesforce Agentforce agents use LLMs to autonomously execute Apex, Flows, and external services. Without guardrails, this creates OWASP LLM Top 10 vulnerabilities:
š“ Excessive Agency ā Agents that modify data without user confirmation
š“ Prompt Injection ā User input that hijacks agent instructions
š“ Privilege Escalation ā Actions running in system context without sharing
š Supply Chain ā Stale API versions that silently skip metadata types
SquireX scans your Agentforce metadata with 61+ SAST rules across 27 categories and integrates directly into your AI coding workflow.
Related MCP server: frogeye-mcp
Quick Start
Add to your AI IDE configuration (Claude Code, Gemini, VS Code, JetBrains):
{
"mcpServers": {
"squirex": {
"command": "npx",
"args": ["-y", "@squirex.dev/mcp-server"],
"env": {
"SQUIREX_PROJECT_DIR": "/path/to/your/salesforce/project"
}
}
}
}That's it. Your AI agent now has access to the Agentforce Capability Scanner.
MCP Surface
Tools (16)
Core Scanning (Primary Value)
Tool | Description |
| Run all 61+ SAST rules against the project. Returns SARIF. |
| Scan a single metadata file ( |
| Run a specific rule (e.g., |
Rule Intelligence
Tool | Description |
| List all 61+ rules with ID, category, severity |
| Deep dive: description + remediation guidance |
| Root-cause analysis for a specific violation |
| Generate a code/metadata fix suggestion |
Apex Testing & Schema
Tool | Description |
| Execute Apex tests locally using the Go interpreter |
| Extract line-level code coverage data |
| Aggregate inferred SObject schema from the codebase |
| Predict merge conflicts between branches |
| Generate SARIF for CI/CD pipeline integration |
Testing Center Bridge
Tool | Description |
| Convert scan violations ā Agentforce DX test YAML for Testing Center |
| Validate DX test spec syntax and schema |
| Push test spec to Salesforce via |
| Get status/results of a Testing Center test run |
Resources (6)
URI | Description |
| Complete 51+ rule catalog |
| Rule detail with remediation |
| Latest scan results (SARIF) |
| Inferred SObject schema |
| Latest Apex test results |
| Latest code coverage |
Prompts (4)
Prompt | What It Does |
| Full 51+ rule scan ā prioritized remediation plan |
| Diagnose + fix a specific violation |
| Proactive defense-in-depth review |
| Generate Agentforce DX test YAML specs for Testing Center |
Agentforce Capability Scanner ā 61+ Rules
# | Category | Rules | Severity |
1 | Action Configuration | Mandatory Confirmation, Schema Sync, Privilege Analysis | š“ Critical / š High |
2 | Agent Script Safety | Validation Guards, Transition Integrity, Prompt Injection Defense | š“ Critical / š High |
3 | Grounding Security | Hardcoded Secrets, FLS Masking Alignment | š“ Critical / š High |
4 | Structural Dependency | Planner Completeness, Deactivation Collision, Evaluation Governance | š High / š” Medium |
5 | Flow Security | Flow Context/Silent State/Injection, API Injection, PT Poisoning/Activation | š“ Critical / š High |
6 | Supply Chain Security | API Downgrade, Schema Desync, Managed Package Origin | š High / š” Medium |
7 | Agentic Architecture | Topic Bloat, Skill Semantics, Orphaned Bot Evaluation | š High / š” Medium |
9 | Instruction Integrity | Metadata Instruction Poisoning, Cross-Topic Boundary | š“ Critical / š High |
10 | Operational Reliability | Validation Conflict | š” Medium |
11 | Autonomous Scheduling | Unguarded Scheduled Action, Time-Window Privilege Drift | š“ Critical / š High |
12 | Copilot Studio Config | Memory Poisoning, API Version Drift | š“ Critical |
13 | Data Cloud Grounding | RAG Without Schema Classification | š“ Critical |
14 | Slack Integration | Slack Bot Without DLP Guard | š High |
15 | External Service Security | Certificate Pinning, Dynamic Cloaking RAG | š High |
16 | Custom Permission | Agent Action Without Permission Gate | š High |
17 | Commerce | Idempotency Key, Amount Bounds Check | š“ Critical |
18 | Multi-Agent Orchestration | Compositional Fragment Trap, Sybil Identity | š“ Critical / š High |
19 | Platform Event / CDC | Sub-agent Spawning Loop, CDC Field Filter | š“ Critical / š High |
20ā24 | OWASP LLM Top 10 | Excessive Agency, Data Exfil, SSRF, DoS, MCP Auth | š“ Critical / š High |
30ā34 | Enterprise Graph | PII Graph, Privilege Graph, Blast Radius, MCP Scope, XSS Graph | š“ Critical / š High |
Supported Metadata Types
.genAiFunction-meta.xml.genAiPlugin-meta.xml.genAiPlannerBundle-meta.xml.genAiPromptTemplate-meta.xml.genAiPromptTemplateActv-meta.xml.agentfiles.cls(Apex classes).trigger(Apex triggers).namedCredential-meta.xml.connectedApp-meta.xml.field-meta.xml.aiEvaluationDefinition-meta.xml.flow-meta.xmlschema.jsonsfdx-project.jsonpackage.xmlLWC components (
.jsinlwc/dirs)
How It Works
AI Coding Agent (Claude / Gemini / Copilot)
ā
ā MCP Protocol (stdio)
ā¼
āāāāāāāāāāāāāāāāāāāāāāāāāāā
ā @squirex.dev/mcp-server ā ā This package
ā 16 tools, 6 resources ā
ā 4 prompts ā
āāāāāāāāā¬āāāāāāāāāāāāāāāāāā
ā spawn
ā¼
āāāāāāāāāāāāāāāāāāāāāāāāāāā
ā squirex CLI ā
ā scan / generate-tests ā ā Testing Center bridge
āāāāāāāāā¬āāāāāāāāāāāāāāāāāā
ā JSON IPC
ā¼
āāāāāāāāāāāāāāāāāāāāāāāāāāā āāāāāāāāāāāāāāāāāāāāāāā
ā squireinterp ā ā sf agent test run ā
ā Go Execution Engine ā ā (Salesforce CLI) ā
ā 61+ SAST Rules ā ā ā Testing Center ā
āāāāāāāāāāāāāāāāāāāāāāāāāāā āāāāāāāāāāāāāāāāāāāāāāāThe Testing Center bridge tools delegate to squirex generate-tests, which:
Runs a capability scan (or reads existing SARIF)
Converts violations to Agentforce DX test YAML (all 61+ rules, normalized
AGENTFORCE-X.YIDs)Optionally validates and pushes to the Salesforce Testing Center via
sf agent test run
Requirements
Node.js ā„ 18
squirexCLI installed (or available via npx)A Salesforce project with Agentforce metadata
GitHub App Integration
For automated PR scanning, install the SquireX GitHub App ā one-click setup, 51+ rule scan on every pull request, SARIF in your Security tab.
Plan | Public Repos | Private Repos | Price |
Free | Unlimited | ā | $0 |
Pro | Unlimited | 1 | $49/repo/month |
Enterprise | Unlimited | Unlimited | $299/org/month |
License
Proprietary ā See LICENSE.md
Copyright Ā© 2026 SquireX. All Rights Reserved.
ā” Built by SquireX ā Securing the AI Agent Pipeline
Available Tools
16 toolsanalyze_schemaA
Aggregate and analyze the inferred schema from all Apex classes in the project. Returns SObject definitions, field usage heatmaps, and relationship graphs. Useful for understanding the data model without an org connection.
Inputs: directory, includeHeatmap.
explain_violationA
Given a SARIF violation from a scan, provide root-cause analysis and step-by-step remediation guidance. Pass the full violation object or just the ruleId and file path.
Inputs: ruleId (required), message, filePath, lineNumber.
generate_dx_testsA
Convert Agentforce scan violations into Agentforce DX test specifications (YAML format) compatible with sf agent test run and the Salesforce Testing Center. Delegates to the core SquireX CLI generate-tests command. If no SARIF file is provided, runs a fresh scan first.
Inputs: rules, sarifPath, suiteName, outputPath.
generate_sarif_reportA
Generate a SARIF (Static Analysis Results Interchange Format) report from the Agentforce Capability Scan. SARIF is the standard format for GitHub Security tab, Azure DevOps, and other CI/CD integrations.
Inputs: outputPath.
get_coverageA
Extract code coverage data from the most recent test run. Shows covered/uncovered lines per class.
Inputs: className.
get_rule_detailsA
Get detailed information about a specific Agentforce SAST rule including its description, severity, and remediation guidance.
Inputs: ruleId (required).
get_testing_center_resultsA
Get the status and results of a previously submitted Agentforce Test Run from the Testing Center. Requires an authenticated Salesforce org.
Inputs: targetOrg (required), testRunId (required).
list_scan_rulesA
List all 26 Agentforce Capability Scanner rules with their ID, name, category, severity, and description. Use this to understand the full scope of the security scan.
Inputs: category, severity.
predict_conflictsB
Predict merge conflicts between two branches by analyzing Apex class changes, DML patterns, and metadata dependencies. Helps prevent deployment failures.
Inputs: branches (required), simulate.
push_to_testing_centerA
Push an Agentforce DX test spec to the Salesforce Testing Center. Delegates to squirex generate-tests --push --target-org <org>. Requires an authenticated Salesforce org with sf org login and the Agentforce DX plugin.
Inputs: targetOrg (required), testFilePath (required).
run_testsA
Execute Apex tests locally using the high-fidelity Go interpreter. Returns structured test results with pass/fail status, assertion details, stack traces, and line-level code coverage.
Inputs: file, filter, failFast, parallel.
scan_agentforceA
Run the full Agentforce Capability Scan (26 SAST rules across 9 categories) against the current project. Returns SARIF-structured violations covering action configuration, agent script safety, grounding security, structural dependencies, flow/prompt template security, supply chain, agentic architecture, instruction integrity, and operational reliability.
Inputs: directory, baseBranch.
scan_agentforce_fileA
Scan a specific Agentforce metadata file for capability violations. Supports .genAiFunction-meta.xml, .genAiPlugin-meta.xml, .genAiPlannerBundle-meta.xml, .genAiPromptTemplate-meta.xml, .agent files, .cls (Apex), .trigger, and schema.json files.
Inputs: filePath (required).
scan_agentforce_ruleA
Run a single Agentforce SAST rule against the project. Useful for focused analysis. Rule IDs follow the pattern AGENTFORCE-X.Y (e.g., AGENTFORCE-1.1 for Mandatory User Confirmation).
Inputs: ruleId (required).
suggest_fixA
Given a specific Agentforce violation (rule ID + file), generate a suggested metadata or code fix. Returns a diff-style suggestion.
Inputs: ruleId (required), filePath (required), violationContext.
validate_dx_testsA
Validate a generated Agentforce DX test spec without executing it. Delegates to squirex generate-tests --validate. Requires Salesforce CLI with Agentforce DX support.
Inputs: testFilePath (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelogā¦
TDQS
Scored across 16 tools
Most tools have clear, distinct scopes (scan whole project vs. specific file vs. specific rule; explain vs. suggest fix). The only minor overlap is between scan_agentforce and generate_sarif_report, both of which output SARIF, which could cause an agent to pick the wrong one when a report artifact is needed.
All tool names follow a consistent verb_noun pattern with lowercase and underscores (scan_agentforce, list_scan_rules, generate_sarif_report, push_to_testing_center). The verbs are predictable and clearly indicate the action, making the set easy to navigate.
With 16 tools, the server is slightly over the typical 3-15 well-scoped range, but each tool serves a distinct purpose within the two main workflows (scanning and testing). The count feels justified and not bloated, so it earns a 4 rather than a 3.
The tool surface covers the full scanning lifecycle (run scans, list rules, get details, generate reports, explain and suggest fixes) and the testing workflow (run tests, get coverage, generate/validate/push test specs, retrieve results). Minor gaps include no direct way to apply suggested fixes or retrieve historical scan results, but these are workable.
Maintenance
Related MCP Connectors
Zero-config MCP security scanner for AI-generated apps. 25K+ vulnerability patterns.
Zero-log MCP server with 49 tools for AI guardrails, secret scanning, MCP linting & EU compliance.
MCP server for secureFlows: token-free URL builders and integration-linting tools for AI agents.
MCP server: static security scanner for MCP servers, agent skills & plugins. 17 attack patterns.
Related MCP Servers
- AlicenseBqualityDmaintenanceAn MCP server that integrates SAST, DAST, and SCA security tools to enable AI-driven vulnerability scanning and automated security reporting. It allows AI assistants to execute and analyze results from tools like Semgrep, OWASP ZAP, and Trivy within a DevSecOps workflow.6MIT

frogeye-mcpofficial
FlicenseNot gradedqualityDmaintenanceAI-powered security vulnerability detection for MCP-compatible agents, offering real-time scanning of code against a knowledge graph of 24,000+ vulnerability patterns.-- AlicenseCqualityDmaintenanceEnterprise-grade MCP server for static application security testing with multi-tool integration, compliance verification, AI-powered remediation, and multi-tenant management.2310MIT
- AlicenseNot gradedqualityDmaintenanceMCP server that securely reviews AI-generated code for vulnerabilities such as SQL injection, command injection, and hardcoded credentials.12Apache 2.0