> [!NOTE] Context > First batch under &23939 (MCP alias map: close gaps from observed usage). Source: the same 60-day telemetry pull behind the [agent tool selection discovery page](https://mcp-tool-selection-661d57.gitlab.io/), window 2026-08-06 to 2026-10-04. ## Summary 78 distinct names arrived at the MCP server in this window that resolved to neither a tool nor an existing alias, each matched here to the real shipped tool the agent was reaching for. 67 are safe to alias now: the guessed name and the shipped tool are the same capability, just different wording, verb convention, or casing. The other 11 are held back below, because aliasing them would silently assume a capability match that isn't established — those need a product decision, not a redirect. ## Recommended additions (67) Grouped by target tool, each guess's distinct-user count in parens. - **get_pipeline_jobs** ← `list_pipeline_jobs` (13u), `list_jobs` (2u), `list_project_jobs` (1u) - **get_merge_request_diffs** ← `get_merge_request_changes` (12u), `get_merge_request_diff` (4u) - **get_repository_file** ← `get_file_contents` (11u), `get_file` (10u), `get_project_file` (1u), `read_file` (1u) - **get_user** ← `get_current_user` (7u), `whoami` (1u), `gl_get_user` (1u) - **get_merge_request_notes** ← `list_merge_request_notes` (7u), `list_merge_request_discussions` (6u), `get_merge_request_discussions` (5u), `mr_discussions` (2u), `discussion_list` (1u) - **list_repository_tree** ← `get_repository_tree` (6u), `list_project_files` (1u), `list_repository_files` (1u), `List_repository_tree` (1u, casing) - **list_merge_requests** ← `search_merge_requests` (6u), `get_merge_requests` (5u), `list_project_merge_requests` (2u) - **list_work_items** ← `list_issues` (5u), `search_workitems` (1u), `search_work_items` (1u), `search_issues` (1u) - **list_projects** ← `search_projects` (4u), `list_all_projects` (1u) - **get_merge_request_pipelines** ← `list_merge_request_pipelines` (4u), `get_mr_pipelines` (1u) - **get_job** ← `get_job_logs` (4u), `get_job_trace` (2u), `get_pipeline_job_log` (1u) — map through the existing `LOG_ALIAS` path (same one `get_job_log` already uses), not bare job metadata - **save_note** ← `create_note` (3u), `create_issue_note` (1u) - **manage_pipeline** ← `retry_job` (3u), `retry_pipeline` (3u), `save_job` (2u) - **list_pipelines** ← `list_project_pipelines` (2u) - **accept_merge_request** ← `merge_merge_request` (2u) - **get_pipeline** ← `get_pipelines` (2u), `gitlab_get_pipeline` (1u) - **add_commit** ← `save_repository_files` (1u) - **get_workitem_notes** ← `get_issue_notes` (1u), `list_issue_discussions` (1u) - **save_pipeline** ← `edit_pipeline` (1u), `create_pipeline` (1u), `update_pipeline` (1u) - **save_merge_request** ← `edit_merge_request` (1u) - **semantic_search** ← `search_repository_code` (1u) — explicitly names "code", unlike the held-back `search_code`/`search_repositories` below - **get_merge_request** ← `get_merge_request_approval_state` (1u), `Direct___get_merge_request` (1u) - **link_work_items** ← `gitlab_link_work_items` (1u) - **search** ← `search_gitlab` (1u), `Search` (1u, casing), `Direct___search` (1u) - **get_commit** ← `get_commit_diff` (1u), `get_repository_commits` (1u) - **create_merge_request** ← `glab_mr_create` (1u) - **list_commits** ← `gitlab_list_commits` (1u) - **save_work_item** ← `update_issue` (1u) — residual from the original 17 ghost-demand cohort, already shipped - **get_mcp_server_version** ← `Gitlab_get_mcp_server_version` (1u, casing), `get_server_version` (1u) - **get_issue** ← `get_issue_by_iid` (1u) - **create_workitem_note** ← `create_work_item_note` (1u) ## Held back — needs a product decision, not a silent alias (11) | Guessed name | Users | Would alias to | Why it's held back | |---|---|---|---| | `search_repositories` | 9 | `semantic_search` | Reads as "find repositories/projects," not code search. Highest-reach holdout here — worth resolving deliberately rather than defaulting. | | `search_code` | 2 | `semantic_search` | Implies literal/grep-style matching; semantic search isn't a guaranteed substitute. | | `search_code_in_project` | 1 | `semantic_search` | Same concern. | | `search_project` | 1 | `semantic_search` | Same concern. | | `grep` | 1 | `semantic_search` | Explicitly wants literal pattern match. Clearest case of the mismatch. | | `get_branch` | 4 | `list_branches` | Asks for one branch; `list_branches` returns all of them. Not the same shape of answer. | | `get_pipeline_bridges` | 1 | `manage_pipeline` | Bridges (downstream/cross-project triggers) are a distinct read; `manage_pipeline` is an action dispatcher, not a getter. | | `get_pipeline_variables` | 1 | `manage_pipeline` | Same concern — a getter mapped onto an action tool. | | `list_pipeline_schedules` | 1 | `list_pipelines` | Schedules are a distinct resource from pipeline runs. | | `ask_duo_agent` | 1 | `start_duo_session` | "Agent"/"flow" vs. "session" isn't confirmed to be the same primitive. | | `trigger_duo_flow` | 1 | `start_duo_session` | Same concern. | ## Acceptance criteria - Each addition above is a registered `tool_alias` resolving to its listed canonical tool; `list_tools` continues to advertise canonical names only. - None of the 11 held-back names are aliased as part of this issue. - Unresolved-name telemetry keeps running so the next window's gap is visible without a fresh manual audit.