Using Alpha Lab

What you need to know (TL;DR)

Alpha Lab = Scored alpha insight feed

from wayfinder_paths.core.clients.AlphaLabClient import ALPHA_LAB_CLIENT

# Search all insights (sorted by score)
await ALPHA_LAB_CLIENT.search()

# Filter by type
await ALPHA_LAB_CLIENT.search(scan_type="twitter_post", min_score=0.7)

# Text search
await ALPHA_LAB_CLIENT.search(search="ETH funding", limit=10)

# List available scan types
await ALPHA_LAB_CLIENT.get_types()

MCP quick access:

research_search_alpha(limit="20")                                                  # Top 20 insights
research_search_alpha(scan_type="twitter_post", limit="10")                        # Top 10 tweets
research_search_alpha(query="ETH", limit="10")                                     # Search "ETH"
research_search_alpha(created_after="2026-03-06T00:00:00Z", limit="20")            # Today's insights
research_search_alpha(sort="-created", limit="20")                                 # Newest insights
research_get_alpha_types()                                                         # List scan types

Tool args: research_search_alpha(query, scan_type, created_after, created_before, sort, limit)

• query: text search, _ for none (default _)
• scan_type: filter by type, all for none (default all)
• created_after / created_before: ISO 8601 datetime bounds, _ to skip (default _)
• sort: -insightfulness_score (default) or -created for newest first
• limit: max results (default "20", max "200")

Critical gotchas:

• Client returns data directly (not tuples) — data = await ALPHA_LAB_CLIENT.search()
• Scores are 0-1 floats (1 = most insightful)
• Default sort is score descending — all-time top scorers, NOT newest. For "latest/fresh alpha" pass sort="-created" (or a created_after bound), or recent items stay invisible unless they crack the all-time top N and the feed looks stale
• Max 200 results per call; use offset for pagination (Python client only)

When to use

• "What's the latest alpha?" / "Show me today's insights"
• "Any interesting tweets about ETH?"
• "What chain flows are happening?"
• "What are the highest-scored insights?"

Alpha Lab is self-contained for alpha requests. It already includes delta_lab_top_apy and delta_lab_best_delta_neutral scan types, so do NOT also query Delta Lab separately. Only use Delta Lab directly when the user asks for raw rates, historical timeseries, or detailed screening — not for "today's alpha".

How to use

• rules/what-is-alpha-lab.md - What Alpha Lab is and its scan types
• rules/high-value-reads.md - Core queries and MCP URIs
• rules/response-structures.md - Response shapes
• rules/gotchas.md - Common mistakes