Dashboard
Agent running…
👤
Total KOLs
—
discovered
🔥
Hot Tier
—
score ≥ 70
🌤
Warm Tier
—
score 45–70
❄️
Cold Tier
—
score < 45
🔗
With Referral
—
affiliate links
⚡
Mindo Mentions
—
mention Mindo
⏳
Pending Review
—
awaiting decision
✓
Approved
—
approved by you
KOL Pipeline
—
Discovered
100%
—
Pending Review
—
Approved
—
Contacted
—
Signed Up
Agent Overview
Loading agents…
KOLs Discovered — Last 14 Days
Quick Actions
✅
Review Queue
Approve or reject KOLs
🗂
KOL Database
Browse all influencers
🤖
Agents
Manage & run agents
📥
Export CSV
Download database
🔍
HandleFollowersER ScoreTierPerformanceWeb3 FlagsCRMDiscovered
Loading…
Run all active agents:
Idle
Agents Click a card to view run history, KOLs, and learning log
Loading agents…
✓ Saved
Filters
Keywords
Exclusions
Scoring
Claude Prompt
Pre-qualification thresholds — accounts that fail any of these are rejected before Claude analysis.
Reject accounts below this follower count
Followers ÷ Following minimum (e.g. 2.0)
Reject accounts younger than this
Minimum Sorsa audience quality score
Reject if Sorsa bot % exceeds this
Maximum new accounts processed each pipeline run
Each group contains search queries sent to Sorsa. The agent cycles through 3 groups per run. Edit queries (one per line) or group names.
Words excluded from every search query. One per line — leading minus is added automatically (e.g. "exchange" becomes -exchange).
If any of these phrases appear in a user's bio, they are immediately rejected. One per line. Case-insensitive substring match.
If a username contains any of these substrings, the account is rejected. One per line. Case-insensitive.
Tune how the composite score (0–100) is calculated. Weights define each signal's contribution to the base score. Bonuses are flat points added on top.
Proven referral/campaign experience
Influence power & crypto-native reach
Audience quality indicator
Content depth and focus
Quality signal (secondary)
Credibility baseline
Account has proven referral link history
Already a Mindo user / partner
Aware of Mindo but not confirmed active
Score ≥ this → Hot tier
Score ≥ this → Warm tier (else Cold)
Raw Sorsa score treated as 100%
ER % treated as 100 in scoring
Override the built-in Claude system prompt. Leave blank to use the built-in. Changes affect all future runs — test carefully.
✓
Run complete
Agent
Active
—
🔥 Hot
—
🌤 Warm
—
❄ Cold
—
Total KOLs
Loading…
Filters
Keywords
Exclusions
Scoring
Schedule
Claude Prompt
Learning Log
Pre-qualification thresholds — accounts failing any of these are rejected before Claude analysis.
Reject accounts below this count
Followers ÷ Following minimum
Reject accounts younger than this
Minimum Sorsa audience quality score
Reject if Sorsa bot % exceeds this
Appends a Sorsa search operator to every query — only tweets containing this media type will be discovered.
How often the scheduler auto-triggers this agent. Choose "Manual" to run it only on demand.
Maximum number of new KOL accounts evaluated and scored per run. Higher values are more thorough but take longer.
Accounts scored & stored per run (default 25)
Search queries sent to Sorsa API. The agent cycles through groups per run — each group can have multiple query lines.
Words excluded from every search query. One per line — leading minus is added automatically.
Accounts with any of these phrases in their bio are immediately rejected. One per line, case-insensitive.
Reject accounts whose username contains any of these substrings. One per line.
Score Weights values as % — should sum to 100% 100%
Proven referral/campaign experience
Influence power & crypto-native reach
Audience quality indicator
Content depth and focus
Quality signal
Credibility baseline
Score Bonuses flat points added after weighted sum
Proven referral link history
Already a Mindo user / partner
Aware of Mindo, not confirmed active
Tier Thresholds
Score ≥ this → Hot tier
Score ≥ this → Warm (else Cold)
Normalization Ceilings raw values mapped to 0–100
Raw Sorsa score treated as 100%
ER % treated as 100 in scoring
Override the built-in Claude system prompt. Leave blank to use the default. Test changes carefully — this affects all future runs for this agent.
Claude receives each candidate's full profile and decides approve / borderline / reject with a score 1–10 and a reason.
The agent automatically adjusts its search queries when runs return 0 results, and re-tunes scoring thresholds as you approve and reject KOLs. This log shows every self-calibration event.