traceAI

What does this agent do, and when should you use it?

traceAI, from the future-agi/traceAI repository (Apache 2.0), is an open-source tracing framework built natively on OpenTelemetry for observing LLM calls, agents, tools, and retrieval in AI applications. It ships as libraries/SDKs in five languages — Python, TypeScript/JavaScript, Java, Go, and C# — with 79 integration families including OpenAI, Anthropic, LangChain/LangGraph, LlamaIndex, Pinecone, and MCP. Once instrumented, model calls, retrieval, and tool execution are connected via OpenTelemetry spans and context, carrying semantic attributes like gen_ai.request.model and input.value, and exported through standard OTLP to your own backend or the Future AGI platform. Python and TypeScript expose privacy environment variables such as FI_HIDE_INPUTS to control what leaves your application. The core pitch: keep your application code and observability stack, and instrument only at the SDK boundary to get request-level traces.

traceAI intercepts supported SDK calls through per-language instrumentors, client wrappers, or middleware and emits OpenTelemetry spans. The Python path: pip install fi-instrumentation-otel and an integration package (e.g., traceai-openai), call register() for a tracer provider, then OpenAIInstrumentor().instrument(tracer_provider=provider) auto-patches the client so each chat.completions.create produces nested agent/tool/chain/LLM spans with model, input/output, and token usage. TypeScript uses register() from @traceai/fi-core with OpenAIInstrumentation from @traceai/openai (manuallyInstrument(OpenAI) for ESM); Java provides wrapper clients such as TracedOpenAIClient; Go offers option.WithMiddleware(traceaiopenai.Middleware(...)); C# exposes manual span APIs (tracer.Chain(...)). Traces export via OTLP to Future AGI or any OpenTelemetry backend you configure.

1. Platform teams adding LLM call tracing to an existing OpenAI application without changing business code
2. Teams on agent frameworks like LangChain, CrewAI, or Pydantic AI debugging tool and retrieval chains across multi-span requests
3. RAG developers tracing retrieval pipelines built on LlamaIndex, Pinecone, Qdrant, or pgvector
4. Teams collecting AI telemetry into their own OTLP backend, reusing an existing OpenTelemetry pipeline instead of Future AGI
5. Compliance-sensitive teams masking inputs, outputs, images, and embedding vectors via FI_HIDE_INPUTS-style variables

How do you install or deploy this agent?

pip install fi-instrumentation-otel traceai-openai

The Python quickstart targets Python 3.10–3.13. Other languages:

npm install @traceai/fi-core @traceai/openai @opentelemetry/instrumentation openai@5

<dependency>
  <groupId>com.futureagi</groupId>
  <artifactId>traceai-java-openai</artifactId>
  <version>1.0.0</version>
</dependency>

go get github.com/future-agi/traceAI/go/openai@latest
go get go.opentelemetry.io/otel/exporters/otlp/otlptrace/[email protected]

dotnet add package fi-instrumentation-otel

How do you use this agent?

Set credentials, then run your instrumented application:

export FI_API_KEY="your-futureagi-api-key"
export FI_SECRET_KEY="your-futureagi-secret-key"
export OPENAI_API_KEY="your-openai-api-key"
export OPENAI_MODEL="gpt-4.1"

import os
from openai import OpenAI
from fi_instrumentation import register
from fi_instrumentation.fi_types import ProjectType
from traceai_openai import OpenAIInstrumentor

provider = register(project_name="traceai-quickstart", project_type=ProjectType.OBSERVE)
OpenAIInstrumentor().instrument(tracer_provider=provider)
try:
    response = OpenAI().chat.completions.create(
        model=os.environ["OPENAI_MODEL"],
        messages=[{"role": "user", "content": "Say hello in one word."}],
    )
    print(response.choices[0].message.content)
finally:
    provider.shutdown()

After a call finishes, open the traceai-quickstart project in Future AGI's Observe; flush or shut down the provider before short-lived processes exit.

What are this agent's strengths and limitations?

How does this agent compare with similar options?

Key facts side by side with the most closely related agents.

How does FollowAgents rate this agent?

Recent releases

Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.

1. traceai-google-genai-live-v0.1.0 traceAI-google-genai-live 0.1.0 2026-11-02
   Release notes excerpt

   - New package: tracing for Gemini Live sessions opened with client.aio.live.connect on the Gemini Developer API (google-genai >=2.28,<3; tested on 2.31.0 and 2.28.0 with gemini-3.8-live). Gemini Live is a preview API: a field rename upstream is a floor and fixture update here, not a silent fallback. google-genai 3.0 removes Live.send / Live.start_stream, which this package never wrapped; pin below 3. - One trace per session: setup, user turns, model turns (model, tokens, transcription), tool …

   View on GitHub ↗
2. traceai-minimax-v0.1.0 traceAI-minimax 0.1.0 2026-11-01
   Release notes excerpt

   _This version has no CHANGELOG entry; see the commits since the previous tag._ Install: pip install traceAI-minimax==0.1.0

   View on GitHub ↗
3. traceai-agentscope-v0.1.0 traceAI-agentscope 0.1.0 2026-11-01
   Release notes excerpt

   _This version has no CHANGELOG entry; see the commits since the previous tag._ Install: pip install traceAI-agentscope==0.1.0

   View on GitHub ↗

FAQ