AgentPulse SDK
Zero-config observability for AI agents. Two lines of code → full LLM call tracking with costs, latency, and traces.
Quick Start
pip install agentpulse
import agentpulse
agentpulse.init() # That's it — all LLM calls are now tracked
import openai
client = openai.OpenAI()
resp = client.chat.completions.create(
model="gpt-4o",
messages=[{"role": "user", "content": "Hello"}]
)
# → Dashboard shows: model, tokens, cost, latency — automatically
What Gets Captured
With just agentpulse.init(), the SDK automatically captures:
- LLM calls — model, tokens, latency, cost (OpenAI, Anthropic, LiteLLM)
- Errors — failed API calls with error details
- Streaming — full metrics even for streamed responses
- Cost estimates — built-in pricing for 30+ models
Structured Tracing
For richer observability, add sessions, agents, and tasks:
import agentpulse
ap = agentpulse.init()
# Group related calls into sessions
with ap.session("daily-email-check") as s:
result = process_emails()
s.log("Processed 5 emails")
s.set_result("success")
# Decorate agents and tasks
@agentpulse.agent(name="email-processor")
class EmailAgent:
@agentpulse.task(name="classify")
def classify(self, email):
return call_llm(email)
# Monitor cron jobs
with ap.cron("nightly-cleanup") as c:
do_cleanup()
# Auto-captures: start, end, duration, success/failure
Manual Events
ap.event("memory_snapshot", {"file": "MEMORY.md", "size_kb": 142})
ap.metric("queue_depth", 23)
ap.alert("Cost spike", severity="warning", details="$5.20 in last hour")
Configuration
All via init() kwargs or environment variables:
agentpulse.init(
api_key="ap_...", # or AGENTPULSE_API_KEY
agent_name="my-agent", # or AGENTPULSE_AGENT (default: hostname)
endpoint="https://...", # or AGENTPULSE_ENDPOINT
enabled=True, # or AGENTPULSE_ENABLED (kill switch)
capture_messages=False, # or AGENTPULSE_CAPTURE_MESSAGES (privacy)
auto_patch=True, # or AGENTPULSE_AUTO_PATCH
debug=False, # or AGENTPULSE_DEBUG
flush_interval=5.0, # seconds between flushes
max_queue_size=10_000, # max buffered events
)
CLI
agentpulse status # Check config and connectivity
agentpulse test # Send a test event
agentpulse costs # Print the built-in cost table
Design Principles
- Zero dependencies — stdlib only (
urllib,json,threading,contextvars) - Never blocks — all reporting is fire-and-forget via background thread
- <1ms overhead — just a timestamp + queue append per LLM call
- Privacy by default — prompt/response content not captured unless opted in
- Unpatchable —
shutdown()restores all original library methods
License
MIT
Metadata
Release files for agentpulse-sdk 0.2.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| agentpulse_sdk-0.2.0.tar.gz | 19.4 kB | Without provenance | Details for agentpulse_sdk-0.2.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| agentpulse_sdk-0.2.0-py3-none-any.whl | Python 3 | none | any | Details for agentpulse_sdk-0.2.0-py3-none-any.whl |
Total release size: 43.3 kB
Release files / agentpulse_sdk-0.2.0.tar.gz
| Download URL | agentpulse_sdk-0.2.0.tar.gz |
|---|---|
| Size | 19.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
b6b5f82b25167527b38516f58eda7be6fe4fc40c316f6383bd63ca40855e60b1
|
|
BLAKE2b-256 checksum How to use checksums |
359a6c98a27e7307f22addf43d5484eef279a1c4be877bbda272ef755d44147f
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.2
|
Release files / agentpulse_sdk-0.2.0-py3-none-any.whl
| Download URL | agentpulse_sdk-0.2.0-py3-none-any.whl |
|---|---|
| Size | 23.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
fb44b5d3d3259cca3d8e8362078d1dc2e42234f4941b5ffb10998a5e7b1f8972
|
|
BLAKE2b-256 checksum How to use checksums |
dd5b4902e2e3e5d452d060bcbdbe49a2c29591f17d5f3874adb9b9571370d217
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.14.2
|
Security for agentpulse-sdk 0.2.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.