Skip to main content

AgentBill CrewAI Integration

OpenTelemetry-based crew tracker for automatically tracking and billing CrewAI agent usage.

Installation

pip install agentbill-crewai

This will automatically install the required dependencies:

  • crewai
  • agentbill-langchain (CrewAI uses LangChain callbacks under the hood)

Quick Start

from agentbill_crewai import track_crew
from crewai import Agent, Task, Crew
from langchain_openai import ChatOpenAI

# 1. Initialize LLM
llm = ChatOpenAI(model="gpt-4o-mini")

# 2. Create agents
researcher = Agent(
    role="Research Analyst",
    goal="Find and analyze data",
    backstory="Expert researcher with attention to detail",
    llm=llm
)

writer = Agent(
    role="Content Writer",
    goal="Write engaging content",
    backstory="Creative writer with storytelling skills",
    llm=llm
)

# 3. Create tasks
research_task = Task(
    description="Research the topic: {topic}",
    agent=researcher,
    expected_output="Comprehensive research findings"
)

writing_task = Task(
    description="Write an article based on the research",
    agent=writer,
    expected_output="Well-written article"
)

# 4. Create crew
crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, writing_task]
)

# 5. Run with AgentBill tracking!
result = track_crew(
    crew=crew,
    inputs={"topic": "AI in healthcare"},
    agentbill_config={
        "api_key": "agb_your_api_key_here",
        "base_url": "https://api.agentbill.io",
        "customer_id": "customer-123",
        "debug": True
    }
)

print(result)

# ✅ Automatically captured:
# - All agent LLM calls
# - Token usage per agent
# - Task execution times
# - Total crew cost
# - Agent-level profitability

Features

  • ✅ Zero-config instrumentation - Just wrap with track_crew()
  • ✅ Agent-level tracking - Track each agent's LLM usage
  • ✅ Task-level metrics - Measure task execution time
  • ✅ Multi-agent support - Track complex multi-agent workflows
  • ✅ Cost calculation - Auto-calculates costs per agent
  • ✅ Crew profitability - Compare crew costs vs revenue
  • ✅ OpenTelemetry compatible - Standard observability

Advanced Usage

Track Revenue Per Crew

result = track_crew(
    crew=crew,
    inputs={"topic": "AI trends"},
    agentbill_config={
        "api_key": "agb_...",
        "base_url": "https://...",
        "customer_id": "customer-123"
    },
    revenue=5.00,  # What you charged for this crew execution
    revenue_metadata={
        "subscription": "enterprise",
        "feature": "research_crew"
    }
)

Use with Custom LLMs

from langchain_anthropic import ChatAnthropic

# Works with any LangChain-compatible LLM
anthropic_llm = ChatAnthropic(model="claude-3-5-sonnet-20241022")

agent = Agent(
    role="Analyst",
    goal="Analyze data",
    backstory="Expert analyst",
    llm=anthropic_llm  # CrewAI auto-tracks this!
)

Sequential vs Parallel Crews

# Sequential crew (default)
sequential_crew = Crew(
    agents=[agent1, agent2],
    tasks=[task1, task2],
    process=Process.sequential  # Tasks run one after another
)

# Parallel crew
from crewai import Process

parallel_crew = Crew(
    agents=[agent1, agent2],
    tasks=[task1, task2],
    process=Process.parallel  # Tasks run concurrently
)

# Both tracked automatically!
track_crew(sequential_crew, {...})
track_crew(parallel_crew, {...})

Hierarchical Crews

# Manager agent delegates to worker agents
manager = Agent(
    role="Project Manager",
    goal="Coordinate the team",
    backstory="Experienced manager",
    llm=llm
)

crew = Crew(
    agents=[manager, worker1, worker2],
    tasks=[task1, task2],
    process=Process.hierarchical,  # Manager delegates
    manager_llm=llm
)

# All agent interactions tracked!
result = track_crew(crew, inputs={...}, agentbill_config={...})

Configuration

agentbill_config = {
    "api_key": "agb_...",           # Required - get from dashboard
    "base_url": "https://...",      # Required - your AgentBill instance
    "customer_id": "customer-123",  # Optional - for multi-tenant apps
    "account_id": "account-456",    # Optional - for account-level tracking
    "debug": True,                  # Optional - enable debug logging
    "batch_size": 10,               # Optional - batch signals before sending
    "flush_interval": 5.0           # Optional - flush interval in seconds
}

How It Works

The crew tracker wraps CrewAI execution:

  1. Inject Callback - Adds AgentBill callback to all agents' LLMs
  2. Track Agents - Monitors each agent's LLM calls
  3. Track Tasks - Measures task execution time
  4. Calculate Costs - Sums up all agent costs
  5. Send Signals - Sends data to AgentBill via the unified OTEL pipeline

All agent interactions are automatically captured without code changes.

Real-World Example: Research Crew

from agentbill_crewai import track_crew
from crewai import Agent, Task, Crew
from crewai_tools import SerperDevTool
from langchain_openai import ChatOpenAI

# Tools
search_tool = SerperDevTool()
llm = ChatOpenAI(model="gpt-4o-mini")

# Agents
researcher = Agent(
    role="Senior Research Analyst",
    goal="Discover cutting-edge developments in {topic}",
    backstory="Veteran researcher with 10+ years experience",
    tools=[search_tool],
    llm=llm
)

analyst = Agent(
    role="Data Analyst",
    goal="Analyze research findings and extract insights",
    backstory="Expert at data analysis and pattern recognition",
    llm=llm
)

writer = Agent(
    role="Content Writer",
    goal="Create compelling content from insights",
    backstory="Award-winning writer with storytelling expertise",
    llm=llm
)

# Tasks
research_task = Task(
    description="Research {topic} and compile findings",
    agent=researcher,
    expected_output="Comprehensive research report"
)

analysis_task = Task(
    description="Analyze research and identify key insights",
    agent=analyst,
    expected_output="Detailed analysis with insights"
)

writing_task = Task(
    description="Write engaging article from analysis",
    agent=writer,
    expected_output="Publication-ready article"
)

# Crew
research_crew = Crew(
    agents=[researcher, analyst, writer],
    tasks=[research_task, analysis_task, writing_task],
    verbose=True
)

# Execute with tracking
result = track_crew(
    crew=research_crew,
    inputs={"topic": "Quantum Computing in Drug Discovery"},
    agentbill_config={
        "api_key": "agb_your_key",
        "base_url": "https://api.agentbill.io",
        "customer_id": "pharma-corp-123"
    },
    revenue=50.00,  # What you charged for this research
    revenue_metadata={
        "client": "PharmaCorp",
        "project": "drug_discovery_research"
    }
)

print("Article:", result)

# ✅ Dashboard shows:
# - Cost per agent (researcher, analyst, writer)
# - Total crew cost
# - Revenue ($50)
# - Net margin (revenue - cost)
# - Agent efficiency metrics

Troubleshooting

Not seeing agent data?

  1. Ensure CrewAI agents have LLMs assigned
  2. Check API key is correct
  3. Enable debug: True to see logs
  4. Verify crew is actually running (not just created)

Missing token counts?

  • Some LLMs don't return usage data
  • OpenAI and Anthropic provide accurate counts
  • Local models may need manual instrumentation

Multiple crews running?

Each track_crew() call is independent - perfect for parallel crew execution!

Idempotent Ingestion (X-Request-Id)

Every OTLP batch this SDK exports to /functions/v1/otel-collector carries an X-Request-Id: <uuid4> header (auto-generated per export, reused across transport-level retries within that export). The collector deduplicates on (company_id, request_id), so a transient 5xx / network retry will not double-write WAL rows, spans, or billing events.

  • No configuration required — the header is emitted automatically by the v7.x line of this SDK.
  • Server fallback: clients that cannot set custom headers may send {"requestId": "..."} at the JSON envelope root.
  • Native OTLP exporters: the standard OTLP-HTTP exporter keeps headers static across batches, so per-batch rotation is not yet wired on that path (PHASE3-R3-SDK-REQUESTID-NATIVE — tracked); the collector still dedupes those payloads on the natural key (company_id, trace_id, span_id).

See the OpenTelemetry Integration docs for the full request-id contract.

License

MIT

Metadata

Release files for agentbill-py-crewai 11.0.2

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

Source distribution (sdist)

Source distribution for agentbill-py-crewai 11.0.2
File Size Uploaded Provenance Actions
agentbill_py_crewai-11.0.2.tar.gz 32.0 kB Without provenance Details for agentbill_py_crewai-11.0.2.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for agentbill-py-crewai 11.0.2
File Interpreter ABI Platform Actions
agentbill_py_crewai-11.0.2-py3-none-any.whl Python 3 none any Details for agentbill_py_crewai-11.0.2-py3-none-any.whl

Total release size: 54.7 kB

Release files / agentbill_py_crewai-11.0.2.tar.gz

Download URL agentbill_py_crewai-11.0.2.tar.gz
Size 32.0 kB
Tags Source
SHA-256 checksum
How to use checksums
fe663464d402f9704957e02e135cb5518ff07a9cc53c12034202efd5fcc6cef0
BLAKE2b-256 checksum
How to use checksums
0afce1252fe90b8ff04c336b165ac17257513350deb027b1bd380dbdd8758854
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.11.15

Release files / agentbill_py_crewai-11.0.2-py3-none-any.whl

Download URL agentbill_py_crewai-11.0.2-py3-none-any.whl
Size 22.7 kB
Tags Python 3
SHA-256 checksum
How to use checksums
cf87ac9df01c2561c9f36b83ec420f909bbea4c593b698b0e40202fb7acfea20
BLAKE2b-256 checksum
How to use checksums
89e6aed7cfb9619d55de59c44bc0b40c8b3a6273a4a11ee29ab2d3b54009e0ca
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.2.0 CPython/3.11.15

Security for agentbill-py-crewai 11.0.2

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.

Release history Release notifications | RSS feed

This release

11.0.2 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page