Agent Logs
Stream and analyze agent activity, decisions, and reasoning in real-time.
Overview
The hyperfold agent logs command provides real-time visibility into agent behavior. Watch negotiations unfold, understand pricing decisions, and debug issues as they happen.
Logs are stored in Cloud Logging and retained for 30 days by default. Configure longer retention in your project settings.
Stream Logs
Watch agent activity in real-time:
bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
# Stream live logs from an agent$ hyperfold agent logs sales-bot-01 [2025-12-19 14:30:15] [INFO] Session started: sess_abc123[2025-12-19 14:30:15] [INFO] Buyer: openai_gpt4_buyer | User: user_xyz[2025-12-19 14:30:16] [ACP] Search request: "waterproof running shoes"[2025-12-19 14:30:17] [DATA] Found 8 products (confidence: 0.91)[2025-12-19 14:30:18] [ACP] Quote request: prod_aero_x2, offer=$120[2025-12-19 14:30:19] [CALC] Floor: $92, List: $180, Competitor: $165[2025-12-19 14:30:19] [DECISION] Counter-offer: $162 (10% gold discount)[2025-12-19 14:30:20] [ACP] Response: counter_offer, price=162.00[2025-12-19 14:30:45] [ACP] Accept received: price=162.00[2025-12-19 14:30:46] [PAYMENT] Processing SPT: spt_live_xxx[2025-12-19 14:30:47] [SUCCESS] Order created: order_789[2025-12-19 14:30:47] [INFO] Session complete: revenue=$162.00 # Follow logs (like tail -f)$ hyperfold agent logs sales-bot-01 --follow # Show logs from all agents$ hyperfold agent logs --all --followLog Levels
| Level | Description |
|---|---|
DEBUG | Detailed internal state (development only) |
INFO | Normal operations, session events |
WARN | Unusual conditions, near-limit situations |
ERROR | Failures requiring attention |
Session Logs
View complete logs for a specific negotiation session:
bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
# View logs for a specific session$ hyperfold agent logs sales-bot-01 --session="sess_abc123" SESSION: sess_abc123━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Started: 2025-12-19T14:30:15ZDuration: 32 secondsStatus: completedOutcome: conversion BUYER INFO: Agent: openai_gpt4_buyer User: user_xyz (Gold tier) History: 12 previous orders, $2,450 total CONVERSATION: Turn 1: Search "waterproof running shoes" → 8 products returned Turn 2: Quote prod_aero_x2 @ $120 → Counter: $162 (10% discount applied) Turn 3: Accept $162 → Order created PRICING DECISIONS: List price: $180.00 Buyer offer: $120.00 Floor price: $92.00 Final price: $162.00 Discount: 10% (Gold tier) Margin: 45% PAYMENT: Method: Stripe SPT Amount: $162.00 Status: succeeded Charge ID: ch_xyz123Reasoning Logs
See the agent's chain of thought for pricing decisions:
bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
# Show agent reasoning (chain of thought)$ hyperfold agent logs sales-bot-01 --reasoning [14:30:18] [REASONING] Processing quote request... INPUT CONTEXT: Product: AeroRun X2 ($180 list) Buyer offer: $120 Customer: Gold tier, 12 orders, $2,450 lifetime Inventory: 847 units (overstock) Competitor: $165 (Amazon) THOUGHT PROCESS: 1. Buyer offered $120, which is 33% below list price 2. My floor price is $92 (cost $80 × 1.15 margin) 3. The offer is above floor, so it's technically acceptable 4. However, I can do better for my margin 5. Customer is Gold tier → 10% discount applies 6. Competitor is at $165, so I should stay below that 7. Optimal price: $162 (10% off, beats competitor by $3) DECISION: counter_offer PRICE: $162.00 REASONING TO BUYER: "I can do $162 for you as a Gold member - that's 10% off and beats Amazon's price." # Verbose reasoning with all context$ hyperfold agent logs sales-bot-01 --reasoning --verbose # Reasoning for specific decision types$ hyperfold agent logs sales-bot-01 --reasoning --filter="decision:reject"Reasoning logs contain detailed decision context. Enable with
ENABLE_REASONING_LOG=true in agent configuration. This adds latency and storage costs.Filtering & Search
Find specific logs with powerful filtering:
bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
# Filter by log level$ hyperfold agent logs sales-bot-01 --level=error$ hyperfold agent logs sales-bot-01 --level=warn # Filter by time range$ hyperfold agent logs sales-bot-01 --since="1h"$ hyperfold agent logs sales-bot-01 --since="2025-12-19T10:00:00Z"$ hyperfold agent logs sales-bot-01 --until="2025-12-19T12:00:00Z" # Filter by event type$ hyperfold agent logs sales-bot-01 --filter="type:payment"$ hyperfold agent logs sales-bot-01 --filter="type:acp"$ hyperfold agent logs sales-bot-01 --filter="type:error" # Search log content$ hyperfold agent logs sales-bot-01 --search="floor price"$ hyperfold agent logs sales-bot-01 --search="prod_aero_x2" # Filter by outcome$ hyperfold agent logs sales-bot-01 --filter="outcome:conversion"$ hyperfold agent logs sales-bot-01 --filter="outcome:rejected"$ hyperfold agent logs sales-bot-01 --filter="outcome:abandoned" # Combine filters$ hyperfold agent logs sales-bot-01 \ --since="24h" \ --filter="outcome:rejected" \ --level=infoExport & Analysis
Export logs for offline analysis or external tools:
bash
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
# Export logs to file$ hyperfold agent logs sales-bot-01 --since="24h" --output=logs.json > [Export] Fetching logs from last 24 hours...> [Write] Writing 1,247 log entries to logs.json✓ Exported to logs.json # Export as CSV for analysis$ hyperfold agent logs sales-bot-01 --since="7d" --format=csv --output=logs.csv # Export session summaries$ hyperfold agent logs sales-bot-01 --sessions --since="24h" --output=sessions.json { "sessions": [ { "session_id": "sess_abc123", "started_at": "2025-12-19T14:30:15Z", "duration_seconds": 32, "outcome": "conversion", "revenue": 162.00, "discount_percent": 10, "turns": 3 } ], "summary": { "total_sessions": 127, "conversions": 42, "conversion_rate": 0.33, "total_revenue": 8450.00, "avg_discount": 12.3 }} # Stream to external logging$ hyperfold agent logs sales-bot-01 --follow | tee -a agent.log # Send to BigQuery for analysis$ hyperfold agent logs export --destination=bigquery --dataset=agent_logsFor advanced analytics, set up Operational Dashboards.