Connecting your data

Customer-facing agents that work with live business state need current records and product actions. Context providers connect them to sources such as web search, files, databases, Slack, Google Drive, and MCP servers.

In default mode, context providers expose a source through query_<source> and, where enabled, update_<source>. The query surface can delegate source-specific work to a sub-agent. Set mode=ContextMode.tools when the calling agent should receive the provider's underlying tools directly.

uv pip install "agno[openai,exa]"
export OPENAI_API_KEY="..."
export EXA_API_KEY="..."

import asyncio

from agno.agent import Agent
from agno.context.web import ExaBackend, WebContextProvider
from agno.models.openai import OpenAIResponses


async def main() -> None:
    web = WebContextProvider(backend=ExaBackend(), model=OpenAIResponses(id="gpt-5.5"))
    await web.asetup()
    try:
        agent = Agent(
            model=OpenAIResponses(id="gpt-5.5"),
            tools=web.get_tools(),
            instructions=web.instructions() + "\nAlways cite URLs inline.",
        )
        await agent.aprint_response("What's the latest pricing for Anthropic's models?")
    finally:
        await web.aclose()


if __name__ == "__main__":
    asyncio.run(main())

The provider's query_web tool is async, so run this agent with arun() or aprint_response().

Query sources at run time

Context providers call the configured source when the agent invokes their tools. Freshness depends on the source API, backend, and any caching it applies. Providers can return source URIs or synthesized text; links and citations depend on the provider and its output.

Vector retrieval remains useful for indexed corpora. Context providers are useful when a task needs source-specific navigation or current records from an external system.

Behavior	What determines it
Source freshness	The source API, provider backend, and caching
Links and citations	The provider's returned documents and the agent's response
Accessible records	The permissions granted to the provider's credentials

Provider modes

Each provider ships a recommended tool surface. Override mode to trade tool count against direct control.

Need	Mode
Provider-recommended tool surface	`ContextMode.default`
One natural-language query tool backed by a sub-agent	`ContextMode.agent`
Direct control over provider operations	`ContextMode.tools` exposes the underlying tools

from agno.context import ContextMode
from agno.context.web import ExaBackend, WebContextProvider

web = WebContextProvider(backend=ExaBackend(), mode=ContextMode.tools)

Sources

Source	Tools	Backed by
Web	`query_web`	Exa, Parallel, or an MCP search backend
Filesystem	`query_fs`	A local directory of files, read-only
Workspace	`query_workspace`	A filesystem path or repo, read-only
Database	`query_database`, `update_database`	SQLAlchemy read and write engines
Knowledge wiki	`query_wiki`, `update_wiki`	Filesystem, a Git repo, or Notion for durable prose memory
Slack	`query_slack`, `update_slack`	Slack API, reads and writes
Google Drive	`query_gdrive`	A service account or OAuth credentials, read-only
Gmail	`query_gmail`, `update_gmail`	Gmail API with Google credentials; writes are off by default
Google Calendar	`query_calendar`, `update_calendar`	Google Calendar API; writes are off by default
MCP servers	`query_mcp_<slug>`	Any MCP server: Linear, GitHub, Notion

Read-only providers

For providers that support writes, write=False removes the update tool from the default surface. A read-only provider such as the web has no update tool. This setting does not change permissions for other clients or integrations that can access the same source.

Next steps

Task	Guide
Reach users where they work	Interfaces
Persist what the agent learns	Sessions and memory