Benchmarks

Performance characteristics of AgenticTime operations.

Test Environment

• Hardware: Apple M2, 16 GB RAM
• OS: macOS 14
• Rust: 1.77 (stable)
• File: .atime with varying entity counts

Core Operations

Operation	100 entities	1,000 entities	10,000 entities
Open file	0.3 ms	1.2 ms	8.5 ms
Add deadline	0.02 ms	0.02 ms	0.03 ms
List deadlines (all)	0.05 ms	0.4 ms	3.8 ms
List deadlines (filtered)	0.03 ms	0.2 ms	1.5 ms
Update deadline	0.02 ms	0.02 ms	0.03 ms
Remove deadline	0.01 ms	0.01 ms	0.02 ms
Save file	0.5 ms	3.2 ms	28 ms
Conflict detection	0.1 ms	0.8 ms	7.2 ms
Decay query	0.001 ms	0.001 ms	0.001 ms

MCP Tool Latency

End-to-end latency for MCP tool calls (including JSON-RPC overhead):

Tool	p50	p99
time_deadline_add	0.3 ms	1.2 ms
time_deadline_list	0.5 ms	4.5 ms
time_schedule_conflicts	0.9 ms	8.0 ms
time_stats	0.1 ms	0.3 ms
time_decay_query	0.05 ms	0.1 ms

File Size

Entity Count	File Size
100	12 KB
1,000	98 KB
10,000	920 KB
100,000	9.1 MB

Memory Usage

Entity Count	RSS
100	4 MB
1,000	6 MB
10,000	18 MB
100,000	142 MB

Concurrent Access

Scenario	Throughput
Sequential reads	50,000 ops/sec
Sequential writes	12,000 ops/sec
4 concurrent readers	180,000 ops/sec
2 readers + 1 writer	35,000 ops/sec

Comparison

AgenticTime vs. common alternatives for temporal data:

Feature	AgenticTime	SQLite + schema	JSON file
Structured temporal types	Native	Manual	Manual
Conflict detection	Built-in	Query-based	Manual
Decay curves	Built-in	Not available	Manual
File portability	Single file	Single file	Single file
Concurrent access	Lock + merge	WAL	No safety
MCP integration	Native	Manual	Manual
Typical query latency	< 1 ms	1-5 ms	10-100 ms