#779 Unify research_goal + run_goal into a single MCP goal tool (customer no longer picks research vs act)

closed high agents mcp ux Created 2026-06-29 02:35 · Updated 2026-06-29 02:35

Description

Edit
Bad UX: customers had to choose research_goal (delegate-only, no outbound) vs run_goal (act+email, no fan-out). FIXED: merged into ONE tool (run_goal); research_goal + RESEARCH_DSL_TEMPLATE removed from highway-mcp/server.py; agent_research.py deleted. The single agent self-routes from the request: fans broad/multi-part requests out via 'delegate' (ONE approval), acts directly for focused asks, or both; delivers via send_email/telegram ONLY if asked, else returns findings. _BASE_CATALOG=[search,shell,http,email,telegram,delegate], _BASE_APPROVALS=[shell.run,http.request,email.send,spawn_subgoals]. Unified system_prompt designed via workflow (3-lens panel+synthesis). Engine source-of-truth agent_run_goal.py synced; config.ini comments updated. Verified via MCP: broad 'compare 5/4 tools' -> fan-out (1 approval) + report, no email; single-fact -> direct, no delegate; research+email path proven (SMTP). highway-mcp.service restarted.

Comments

Loading comments...

Context

Loading context...

Audit History

View All
Loading audit history...