Feature Area

Integration with external tools

Is your feature request related to a an existing bug? Please link it here.

NA

Describe the solution you'd like

Problem

CrewAI agents that operate on email (support triage, sales follow-ups, executive assistants) typically get raw message text via Gmail APIs or third-party connectors. That's enough to fetch emails, but not enough to reason about decisions, commitments, unresolved threads, tone changes, and repeated patterns across multiple threads.

Proposal

Add two optional tools implemented using CrewAI's BaseTool + args_schema pattern:

IgptEmailAskTool — Natural-language Q&A over a user's full email history

• Use cases: "What are the open commitments with X?", "What did we decide about pricing?", "Summarize unresolved threads this week."
• Output: text or JSON; includes citations back to source messages (thread/message identifiers, excerpt + timestamp for each citation).

IgptEmailSearchTool — Keyword + semantic search across threads with date filters

• Use cases: gather context before drafting, find prior discussions, trace a topic over time.

Implementation sketch

import os
from typing import Type
from crewai.tools import BaseTool
from pydantic import BaseModel, Field


class IgptEmailAskInput(BaseModel):
    question: str = Field(..., description="Natural language question about email history")
    output_format: str = Field(default="text", description="text | json | schema")


class IgptEmailAskTool(BaseTool):
    name: str = "iGPT Email Ask"
    description: str = (
        "Ask questions over a user's email history and get answers with citations "
        "(useful for decisions, commitments, action items, sentiment, and open loops)."
    )
    args_schema: Type[BaseModel] = IgptEmailAskInput

    def _run(self, question: str, output_format: str = "text") -> str:
        from igptai import IGPT
        igpt = IGPT(
            api_key=os.environ["IGPT_API_KEY"],
            user=os.environ["IGPT_API_USER"],
        )
        return str(igpt.recall.ask(
            input=question,
            quality="cef-1-normal",
            output_format=output_format
        ))

Dependencies / configuration

• Python SDK: igptai (MIT licensed) — https://pypi.org/project/igptai/
• Env vars: IGPT_API_KEY, IGPT_API_USER

Why this belongs as a first-class tool

It's a thin, opt-in wrapper that adds "email intelligence" (cross-thread context + grounded citations) without changing CrewAI core behavior unless a developer installs and uses it.

Resources

• SDK docs: https://docs.igpt.ai
• Python SDK: https://github.com/igptai/igptai-python
• LangChain integration (reference): https://pypi.org/project/langchain-igpt/

Happy to submit a PR with the full implementation if there's interest.

Describe alternatives you've considered

The current alternatives for email-aware CrewAI agents are: (1) building a custom Gmail API + RAG pipeline from scratch, which requires handling threading, deduplication, and participant tracking manually, or (2) using raw Gmail connectors that return message text without cross-thread context. Both give agents access to individual emails but not the structured understanding (decisions, commitments, open loops) needed for reliable reasoning.

Additional context

We've already shipped a LangChain integration (langchain-igpt v1.1.0 on PyPI) using the same BaseTool pattern, and a LlamaIndex integration was merged upstream (PR #20674). Happy to follow whatever conventions the CrewAI tools package uses.

Willingness to Contribute

Yes, I'd be happy to submit a pull request