Skip to content

feat: surface MCP asset descriptions and metadata in readable output - #464

Draft
Christian-Vallejos wants to merge 1 commit into
salesforcecli:mainfrom
Christian-Vallejos:t/api-catalog/mcp-asset-descriptions
Draft

feat: surface MCP asset descriptions and metadata in readable output#464
Christian-Vallejos wants to merge 1 commit into
salesforcecli:mainfrom
Christian-Vallejos:t/api-catalog/mcp-asset-descriptions

Conversation

@Christian-Vallejos

@Christian-Vallejos Christian-Vallejos commented Jul 24, 2026

Copy link
Copy Markdown
Contributor

Draft — blocked on a dependency. This needs a @salesforce/agents release that includes the securityWarning field on McpFetchedAsset (forcedotcom/agents#322). Once that publishes, @salesforce/agents must be bumped in this repo (same as #461 did for 1.10.3), then this can go ready.

What

The sf agent mcp commands drop several fields that the Connect API already returns, so a user only sees them via --json. This brings the human-readable output to parity with the API — most importantly the per-tool description and security warning, so a tool can be reviewed before it is approved.

Command Added to readable output
fetch Label, Description, Security Warning (only /fetch populates it)
asset list Label, Description
asset replace Label, Description, Agent Action
create now prints the discovered-asset table (Label/Description/…) instead of only Discovered N asset(s)
list Auth Type, Description
get Auth Type, Description

--json output is unchanged (it always carried the full object).

Why

Before approving an MCP tool as an agent action, a user needs to read its description and any security warning. Today that information is returned by the API but hidden from the default CLI output.

Dependency

Test plan

  • tsc + eslint clean
  • mocha test/commands/agent/mcp/mcp.test.ts — 18/18 passing (added coverage asserting the Description and Security Warning columns render)
  • schema:compare clean after regeneration
  • pre-push build + test gate green

@W-23572634@

@Christian-Vallejos
Christian-Vallejos force-pushed the t/api-catalog/mcp-asset-descriptions branch from 54203ce to 1ef5f77 Compare July 24, 2026 17:08
@W-23572634@

The Connect API already returns descriptions, labels, auth type and (on
fetch) a per-tool security warning, but the human-readable tables dropped
them. This brings `sf agent mcp` output to parity with the API so a user
can review a tool before approving it, without needing --json.

- fetch / asset list / asset replace / create: add Label + Description
  columns (create now prints the discovered-asset table instead of just
  a count).
- fetch: add a Security Warning column (only /fetch populates it).
- list / get: add Auth Type + Description.

--json output is unchanged (it always carried the full object).

Requires @salesforce/agents with the securityWarning field
(forcedotcom/agents#322); schemas regenerated for fetch + create.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@Christian-Vallejos
Christian-Vallejos force-pushed the t/api-catalog/mcp-asset-descriptions branch from 1ef5f77 to a20e285 Compare July 25, 2026 18:49
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant