Named Query Works From API but Fails as MCP Tool
Named Query works as REST. Salesforce hosted MCP fails Tomas integer calls in Claude, ChatGPT, and Copilot. GPTfy MCP does not wrap Named Query API.
Last updated:
4 hrsper week lost toggling between Salesforce and AI tools. MCP eliminates the gap (Harvard Business Review, 2022)
See GPTfy's MCP server bring live Salesforce data to any AI agent
Connect Claude, Cursor, or any MCP-compatible agent to Salesforce records in minutes. No middleware, no data copying.
Your AI Tools Are Flying Blind.
Daily app switches cost 4 hours of productive time per week
Sales reps open Salesforce, copy account details, paste into an AI chat, and get a generic response. Every session starts blind to what the CRM actually shows.
to ship a custom Salesforce-to-AI connector
Engineering teams scope bespoke integrations, negotiate API limits, build OAuth flows, and maintain field mappings as orgs evolve. By the time it ships, the requirements have changed.
of rep time spent on tasks other than selling
Without a live data connection, AI-generated account plans and pipeline summaries are built from whatever the rep pastes in. Stale data, generic output.
Why does Named Query work from the API but fail as an MCP tool?
Workbench against the Named Query API succeeds
Workbench against the Named Query API succeeds with Integer and Date as REST URI query parameters. The same Named Query, exposed as a Salesforce hosted MCP tool, fails Tomas's integer call with "Unexpected error while invoking tool" in Claude, ChatGPT, and Copilot.
GPTfy does not claim an integer/date MCP serialization fix
GPTfy does not claim an integer/date MCP serialization fix. GPTfy MCP does not wrap Named Query API.
Is Named Query the same as Named Credential?
Named Query and Named Credential are different Salesforce objects
No. Named Query and Named Credential are different Salesforce objects. Named Query is the Salesforce query API in Tomas's report.
Does GPTfy wrap Named Query API?
No. GPTfy MCP publishes GPTfy prompts as tools, with 4-layer masking. It does not wrap Named Query API.
Is Copilot as an MCP client the same as the AppSource Copilot agent?
Copilot as an MCP client is not the AppSource Copilot agent
No. Copilot as an MCP client is not the AppSource Copilot agent. Copilot as an MCP client is one of the clients where Tomas's integer call failed.
What does GPTfy MCP publish?
GPTfy prompts, as MCP tools, with 4-layer masking. Not Named Query API. Not Named Credentials.
REST Named Query API: Integer / Date in as REST URI query parameters. Workbench works. What it exposes: Named Query API.
Salesforce hosted MCP tool: Tomas's integer call fails: "Unexpected error while invoking tool" in Claude, ChatGPT, and Copilot. What it exposes: That Named Query, as a tool.
GPTfy MCP: Does not wrap Named Query API. What it exposes: GPTfy prompts as tools, with 4-layer masking.
Why Choose GPTfy MCP Server
GPTfy prompts as MCP tools
GPTfy MCP publishes GPTfy prompts as tools with 4-layer masking. It does not wrap Named Query API.
Named Query is not Named Credential
Named Query and Named Credential are different Salesforce objects. Named Query is the Salesforce query API in Tomas's report.
Does not wrap Named Query API
You configure GPTfy prompts. GPTfy publishes them as MCP tools with 4-layer masking. That is not Named Query API.
Copilot as an MCP client
Copilot as an MCP client is not the AppSource Copilot agent. Copilot as an MCP client is one of the clients where Tomas's integer call failed.
Powerful Capabilities
MCP Server
Publishes all configured GPTfy agent prompts as named MCP tools for any connected AI client.
A2A Protocol
Supports agent-to-agent communication so AI agents in different tools coordinate across your stack.
Live Data Access
Contacts, opportunities, cases, assets, and custom objects pulled from Salesforce at query time.
Account 360 Generation
Automated planning documents built from live Salesforce records, downloadable as Word files.
Native Security
GPTfy's Salesforce-native permission model and PII masking applies to every MCP tool call.
Open Protocol
Built on the open MCP standard, ensuring compatibility with Claude, ChatGPT, and future MCP clients.
Multi-Source Aggregation
Combine Salesforce data with other MCP-connected sources: files, databases, calendars, email.
Client Agnostic
No vendor lock-in. Switch or add AI clients without touching your GPTfy configuration.
Key Takeaways
- Salesforce Named Query API accepts Integer and Date as REST URI query parameters (Workbench works). Salesforce hosted MCP exposes that query as a tool, and Tomas's integer call fails with "Unexpected error while invoking tool" in Claude, ChatGPT, and Copilot.
- GPTfy MCP publishes GPTfy prompts as tools with 4-layer masking and does not wrap Named Query API.
- Named Query is not Named Credential. Copilot as an MCP client is not the AppSource Copilot agent.
- GPTfy does not claim an integer/date MCP serialization fix. GPTfy MCP does not wrap Named Query API.
Frequently Asked Questions
The Named Query API accepts Integer and Date as REST URI query parameters. Workbench works. Salesforce hosted MCP exposes that query as a tool, and Tomas's integer call fails with "Unexpected error while invoking tool" in Claude, ChatGPT, and Copilot.
No. GPTfy does not claim an integer/date MCP serialization fix. GPTfy MCP does not wrap Named Query API.
No.
No.
Connect Your AI Clients to Salesforce
See how GPTfy's MCP Server makes your Salesforce data available to Claude, ChatGPT, and any MCP-enabled client in your workflow.
Explore More Features
GPTfy Agents
Deploy AI agents natively inside Salesforce with full audit trail and org-level security.
Prompt Builder
Configure the agent prompts that GPTfy publishes as MCP tools.
Bring Your Own Model
Connect any AI model to GPTfy via Named Credentials for in-Salesforce inference.
API Data Sources
Pull data from external APIs into Salesforce AI workflows.
GPTfy Security Layer
Org-level permissions and PII masking that apply across all GPTfy features.