Context
packages/loopover-mcp/bin/loopover-mcp.js implements several maintain CLI subcommands that call real REST endpoints but have no corresponding MCP tool on either server:
maintain queue/approve/reject (lines 1763-1787) → /v1/repos/:owner/:repo/agent/pending-actions[/:id/:decision] (src/api/routes.ts:2537,2547). The remote server already mirrors this as loopover_list_pending_actions/loopover_decide_pending_action (src/mcp/server.ts:2349,2360) — the local stdio server has neither.maintain pause/resume (lines 1788-1792) → PUT /v1/repos/:owner/:repo/settings with agentPaused. Mirrored remotely as loopover_set_agent_paused (src/mcp/server.ts:2314, added per its own comment at lines 528-531 noting this exact CLI gap). Missing locally.maintain set-level (lines 1793-1804). Mirrored remotely as loopover_set_action_autonomy (src/mcp/server.ts:2327). Missing locally.maintain precision (lines 1806-1825) → GET /v1/repos/:owner/:repo/gate-precision. Mirrored remotely as loopover_get_gate_precision (src/mcp/server.ts:1789). Missing locally.
Requirements
- Add local
registerStdioTool(...) registrations for loopover_list_pending_actions, loopover_decide_pending_action, loopover_set_agent_paused, loopover_set_action_autonomy, and loopover_get_gate_precision to packages/loopover-mcp/bin/loopover-mcp.js, backed by the same REST endpoints the CLI subcommands already call (reuse the CLI's existing HTTP-client logic rather than duplicating it).
- Do not change the remote server or the existing CLI subcommands' own behavior.
Test Coverage Requirements
Match packages/loopover-mcp's existing test conventions; at least one test per new tool covering success and a representative failure path.
Deliverables
Expected Outcome
An agent using the local MCP server can perform maintainer queue/pause/autonomy/precision actions without shelling out to the CLI.
Links & Resources
packages/loopover-mcp/bin/loopover-mcp.js:1763-1825src/mcp/server.ts:2349,2360,2314,2327,1789src/api/routes.ts:2537,2547
Context
packages/loopover-mcp/bin/loopover-mcp.jsimplements severalmaintainCLI subcommands that call real REST endpoints but have no corresponding MCP tool on either server:maintain queue/approve/reject(lines 1763-1787) →/v1/repos/:owner/:repo/agent/pending-actions[/:id/:decision](src/api/routes.ts:2537,2547). The remote server already mirrors this asloopover_list_pending_actions/loopover_decide_pending_action(src/mcp/server.ts:2349,2360) — the local stdio server has neither.maintain pause/resume(lines 1788-1792) →PUT /v1/repos/:owner/:repo/settingswithagentPaused. Mirrored remotely asloopover_set_agent_paused(src/mcp/server.ts:2314, added per its own comment at lines 528-531 noting this exact CLI gap). Missing locally.maintain set-level(lines 1793-1804). Mirrored remotely asloopover_set_action_autonomy(src/mcp/server.ts:2327). Missing locally.maintain precision(lines 1806-1825) →GET /v1/repos/:owner/:repo/gate-precision. Mirrored remotely asloopover_get_gate_precision(src/mcp/server.ts:1789). Missing locally.Requirements
registerStdioTool(...)registrations forloopover_list_pending_actions,loopover_decide_pending_action,loopover_set_agent_paused,loopover_set_action_autonomy, andloopover_get_gate_precisiontopackages/loopover-mcp/bin/loopover-mcp.js, backed by the same REST endpoints the CLI subcommands already call (reuse the CLI's existing HTTP-client logic rather than duplicating it).Test Coverage Requirements
Match
packages/loopover-mcp's existing test conventions; at least one test per new tool covering success and a representative failure path.Deliverables
Expected Outcome
An agent using the local MCP server can perform maintainer queue/pause/autonomy/precision actions without shelling out to the CLI.
Links & Resources
packages/loopover-mcp/bin/loopover-mcp.js:1763-1825src/mcp/server.ts:2349,2360,2314,2327,1789src/api/routes.ts:2537,2547