Loading ToolBench

Reading the current corpus…

salesforce-mcp MCP server · F · ToolBench

← Browse

salesforce-mcp

github.com/pouyanafisi/salesforce-mcp

MCP Server for interacting with Salesforce instances

<=2025-11-25STDIO16 tools

44/100

Definition Quality38

Protocol Readiness47

Supportability53

Spec posture

Inferred effective spec <=2025-11-25·current baseline 2026-07-28

Static source inference · medium confidence · detected: Logging

Deprecated protocol patterns detected

Summary

16 tools with mixed quality. Naming is verb-based and clear (get_, create_, delete_, assign_, deploy_, retrieve_). Descriptions present but brief (avg ~60 chars, well below 194-char baseline). Critical gap: input schemas not visible in provided source code, cannot verify parameter types, descriptions, or constraints. Tool definitions appear to exist but are not shown in the sample code provided. Error handling and output schemas not documented in visible code.

Tools (16)

sf-assign-permission-setwriteauth50/100

Assign a permission set to a user in a Salesforce org

sf-create-org-snapshotwriteauth50/100

Create a snapshot of a Salesforce org

sf-create-scratch-orgwriteauth50/100

Create a new Salesforce scratch org

sf-delete-orgdestructiveauth50/100

Delete a Salesforce org

sf-deploy-metadatawriteauth50/100

Deploy metadata to a Salesforce org

sf-enable-toolswritesource verified47/100

Enable one or more of the tools the Salesforce MCP server provides.

sf-get-usernameread onlyauthsource verified40/100

Get the username of the current Salesforce org

sf-list-all-orgsread onlyauth42/100

Issues (5)

CRITICALconstrained-input ↗

Input schemas not visible in source code. Cannot verify parameter types, descriptions, or constraints for most tools.

sf-get-usernamesf-resumesf-list-toolssf-list-all-orgssf-test-agentssf-org-open

HIGHtool-description ↗

Descriptions are too brief (avg ~50 chars vs 194-char baseline). Examples: 'Get the username of the current Salesforce org' (52 chars), 'Resume a previous conversation or session' (42 chars). Lack context for LLM tool selection and when to use each tool.

sf-get-usernamesf-resumesf-list-toolssf-list-all-orgssf-test-agents

HIGHresponse-shaper ↗

No output schemas documented. LLMs cannot infer what fields to expect from tool responses, preventing downstream tool chaining and forcing extra discovery calls.

sf-list-all-orgssf-query-orgsf-retrieve-metadatasf-test-apex

Recommendations

Expand tool descriptions to 100-200 chars. Example: 'sf-query-org: Execute a SOQL query against a Salesforce org. Requires read access to the target objects. Returns up to 50 records; use pagination for larger result sets. Org can be a username or alias.'
Document all parameter types and constraints in JSON Schema. For 'org' param: type=string, enum or pattern for valid formats, description of resolution order (alias → username → email).
Add output schemas to all tools. Example for sf-list-all-orgs: {type: 'object', properties: {orgs: {type: 'array', items: {type: 'object', properties: {alias, username, status, isScratchOrg}}}}, total: {type: 'integer'}}.
Implement confirmation flow for destructive tools (sf-delete-org, sf-deploy-metadata). Use MRTR pattern (result: 'input_required') to ask user for confirmation before executing.
Add error recovery guidance. Example: 'If org not found, call sf-list-all-orgs to see available orgs and their aliases.' Include actionable error messages with valid option suggestions.

Spec posture evidence

Inferred effective spec: <=2025-11-25.

Relies on Logging (deprecated) - log to stderr or use OpenTelemetry

Score history

Overall score trend

First recorded score · v2 rubric

44/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	44	<=2025-11-25	v2