send_message, and the write agent never sees history or search tools. Uploads / downloads are off on both.
slack.py
"""
Slack Context Provider
======================
SlackContextProvider exposes two tools to the calling agent:
- `query_<id>(question)` — read the workspace (search, channel
history, threads, user / channel lookups)
- `update_<id>(instruction)` — post a message (resolves channel /
user names, then calls `send_message` / `send_message_thread`)
Separate sub-agents under the hood keep scopes minimal: read agents
never see `send_message`, and the write agent never sees history or
search tools. Uploads / downloads are off on both.
This cookbook always runs the read prompt. If you set
`SLACK_WRITE_CHANNEL` (e.g. `SLACK_WRITE_CHANNEL=#agno-test`), it
also runs a write prompt that posts a hello message there. Without
it, posting is skipped so a casual `python cookbook/12_context/05_slack.py`
never spams a real channel.
Requires:
OPENAI_API_KEY
SLACK_BOT_TOKEN (bot token; xoxb-...)
With scopes: channels:read, users:read; add
chat:write to exercise the write path.
Optional:
SLACK_TOKEN (falls back here if SLACK_BOT_TOKEN isn't set)
SLACK_USER_TOKEN (user token; xoxp-...) for search_messages API
SLACK_WRITE_CHANNEL (e.g. `#agno-test`) — opt in to the write demo
"""
from __future__ import annotations
import asyncio
from agno.agent import Agent
from agno.context.slack import SlackContextProvider
from agno.models.openai import OpenAIResponses
# ---------------------------------------------------------------------------
# Create the provider (token read from SLACK_BOT_TOKEN / SLACK_TOKEN)
# ---------------------------------------------------------------------------
slack = SlackContextProvider(model=OpenAIResponses(id="gpt-5.4-mini"))
# ---------------------------------------------------------------------------
# Create the Agent
# ---------------------------------------------------------------------------
agent = Agent(
model=OpenAIResponses(id="gpt-5.4"),
tools=slack.get_tools(),
instructions=slack.instructions(),
markdown=True,
)
async def main() -> None:
print(f"\nslack.status() = {slack.status()}\n")
# --- Read path (always runs) ---
# CLI runs use bot-token-compatible channel history. Slack interface
# runs include an action_token, so the provider can use assistant search.
read_prompt = (
"Find the 3 most recent messages in the #agents channel."
"For each, author, and a one-line quote."
)
print(f"> {read_prompt}\n")
await agent.aprint_response(read_prompt)
# --- Write path (opt in via env) ---
write_channel = "#agents"
write_prompt = f"Post the message 'Hello from agno.context' to {write_channel}."
print(f"\n> {write_prompt}\n")
await agent.aprint_response(write_prompt)
if __name__ == "__main__":
asyncio.run(main())
Run the Example
1
Set up your virtual environment
uv venv --python 3.12
source .venv/bin/activate
uv venv --python 3.12
.venv\Scripts\activate
2
Install dependencies
uv pip install -U agno openai
3
Export environment variables
export OPENAI_API_KEY="your_openai_api_key_here"
export SLACK_BOT_TOKEN="your_slack_bot_token_here"
export SLACK_TOKEN="your_slack_token_here"
export SLACK_USER_TOKEN="your_slack_user_token_here"
$Env:OPENAI_API_KEY="your_openai_api_key_here"
$Env:SLACK_BOT_TOKEN="your_slack_bot_token_here"
$Env:SLACK_TOKEN="your_slack_token_here"
$Env:SLACK_USER_TOKEN="your_slack_user_token_here"
4
Run the example
Save the code above as
slack.py, then run:python slack.py