Niitaka SDK

AI agent observability and control plane — one import away.

Niitaka instruments your AI agents to capture every LLM call, tool invocation, and decision trace. Connect to the Niitaka dashboard to run experiments, enforce guardrails, and version your agents — without changing your agent code.

Install

pip install niitaka-sdk                    # core
pip install "niitaka-sdk[openai]"          # + OpenAI
pip install "niitaka-sdk[anthropic]"       # + Anthropic
pip install "niitaka-sdk[all]"             # all providers

Quick start

import niitaka
import os

niitaka.configure(
    api_key=os.getenv("NIITAKA_API_KEY"),
    api_url="https://api.niitaka.ai",      # or your self-hosted URL
)

# Instrument your LLM provider (pick one or more)
niitaka.instrument_openai()
niitaka.instrument_anthropic()
niitaka.instrument_gemini()
niitaka.instrument_groq()

# Wrap your agent run in a session
with niitaka.start_session(goal="Summarise this document", agent_id="my-agent"):
    response = client.chat.completions.create(...)  # auto-logged

Every call inside start_session is automatically traced — latency, token usage, cost, errors, and tool calls.

Providers

Provider	Extra	Import
OpenAI	`niitaka-sdk[openai]`	`niitaka.instrument_openai()`
Anthropic	`niitaka-sdk[anthropic]`	`niitaka.instrument_anthropic()`
Google Gemini	`niitaka-sdk[gemini]`	`niitaka.instrument_gemini()`
Groq	`niitaka-sdk[groq]`	`niitaka.instrument_groq()`
LangChain	`niitaka-sdk[langchain]`	`from niitaka import NiitakaCallbackHandler`

LangChain

from niitaka import NiitakaCallbackHandler
import niitaka

with niitaka.start_session(goal="...", agent_id="lc-agent"):
    chain.invoke({"input": "..."}, config={"callbacks": [NiitakaCallbackHandler()]})

Runtime config

Niitaka can push model, temperature, and guardrail config to your agent at runtime — no redeploy needed:

config = niitaka.get_runtime_config(agent_id="my-agent")
# config["llm"]["model"], config["guardrails"]["cost_limit_usd"], ...

Release files for niitaka-sdk 0.1.25

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for niitaka-sdk 0.1.25
File	Size	Uploaded
niitaka_sdk-0.1.25.tar.gz	57.6 kB	Sep 23, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for niitaka-sdk 0.1.25
File	Interpreter	ABI	Platform	Reset
niitaka_sdk-0.1.25-py3-none-any.whl 63.7 kB Sep 23, 2026	Python 3	none	any	Details

Total release size: 121.3 kB

Release files / niitaka_sdk-0.1.25.tar.gz

Download URL	niitaka_sdk-0.1.25.tar.gz
Size	57.6 kB
Tags	Source
SHA-256 checksum How to use checksums	`da9a8adb0b306ae89061f5fe203081d861b8e2f74ecabbf5f34c1072e096aa49`
BLAKE2b-256 checksum How to use checksums	`a194cc56c5035278de17e595cbcbd7240289bbbb45a5aa180fa77d2bd242cf86`
Upload date	Sep 23, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.11.16`

Release files / niitaka_sdk-0.1.25-py3-none-any.whl

Download URL	niitaka_sdk-0.1.25-py3-none-any.whl
Size	63.7 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`a021930beb93264f02c3dc5071aeb4d84513df96ddd009d1560a20753826db41`
BLAKE2b-256 checksum How to use checksums	`8b80faf0e4b98fb026ff08ef604943df81ee6db470fb6348790267f239d29a86`
Upload date	Sep 23, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.11.16`