MCP server URL or name
https://api.salesforce.com/platform/mcp/v1/platform/sobject-reads
How are you connecting to this server?
From a custom connector URL I entered manually
Where does the issue occur?
When auth finishes and redirects back to Claude.ai
When did you last reproduce this?
2026-05-11, ET
Browser and OS
Chrome on macOS
Describe the issue
Connecting Salesforce Hosted MCP (sobject-reads, GA endpoint) as a custom connector on a production org with a Claude Team plan. OAuth flow completes successfully — Salesforce consent screen appears, I click Allow, token is issued and confirmed in Salesforce OAuth Usage — but Claude returns "Authorization with the MCP server failed" immediately after.
Error reference codes: ofid_16dc23ccd13890ec
This matches the pattern reported in #171 and #184, which Anthropic support confirmed is an Anthropic-side issue. Reproduces consistently across multiple attempts. Salesforce External Client App is configured per GA docs with scopes: Access Salesforce hosted MCP servers, refresh_token, api. Callback URL: https://claude.ai/api/mcp/auth_callback. PKCE enabled, IP relaxed, all users can self-authorize.
Error details
Authorization with the MCP server failed. You can check your credentials and permissions. If this persists, share this reference with support: "ofid_16dc23ccd13890ec"
Expected behavior
After completing OAuth and clicking Allow on the Salesforce consent screen, the connector should show as Connected and the sobject-reads tools should be available in conversations.
Additional context
This is a production org on a Claude Team plan. The same issue is documented in #171 and #184, where Anthropic support confirmed it is an Anthropic-side bug. Salesforce has also acknowledged they are working with Anthropic on a Claude-specific fix (forcedotcom/mcp-hosted#12).
Multiple ofid codes from previous attempts today: ofid_199a9306f9bc082b, ofid_bc59966f3e13c906, ofid_16dc23ccd13890ec.
Salesforce OAuth Usage confirms the token is being issued and touched on each attempt, so the issue is in Claude's post-auth MCP handshake, not in the Salesforce configuration.
MCP server URL or name
https://api.salesforce.com/platform/mcp/v1/platform/sobject-reads
How are you connecting to this server?
From a custom connector URL I entered manually
Where does the issue occur?
When auth finishes and redirects back to Claude.ai
When did you last reproduce this?
2026-05-11, ET
Browser and OS
Chrome on macOS
Describe the issue
Connecting Salesforce Hosted MCP (sobject-reads, GA endpoint) as a custom connector on a production org with a Claude Team plan. OAuth flow completes successfully — Salesforce consent screen appears, I click Allow, token is issued and confirmed in Salesforce OAuth Usage — but Claude returns "Authorization with the MCP server failed" immediately after.
Error reference codes: ofid_16dc23ccd13890ec
This matches the pattern reported in #171 and #184, which Anthropic support confirmed is an Anthropic-side issue. Reproduces consistently across multiple attempts. Salesforce External Client App is configured per GA docs with scopes: Access Salesforce hosted MCP servers, refresh_token, api. Callback URL: https://claude.ai/api/mcp/auth_callback. PKCE enabled, IP relaxed, all users can self-authorize.
Error details
Authorization with the MCP server failed. You can check your credentials and permissions. If this persists, share this reference with support: "ofid_16dc23ccd13890ec"Expected behavior
After completing OAuth and clicking Allow on the Salesforce consent screen, the connector should show as Connected and the sobject-reads tools should be available in conversations.
Additional context
This is a production org on a Claude Team plan. The same issue is documented in #171 and #184, where Anthropic support confirmed it is an Anthropic-side bug. Salesforce has also acknowledged they are working with Anthropic on a Claude-specific fix (forcedotcom/mcp-hosted#12).
Multiple ofid codes from previous attempts today: ofid_199a9306f9bc082b, ofid_bc59966f3e13c906, ofid_16dc23ccd13890ec.
Salesforce OAuth Usage confirms the token is being issued and touched on each attempt, so the issue is in Claude's post-auth MCP handshake, not in the Salesforce configuration.