Keenable Search

KeenableSearchToolkit gives an agent two related tools powered by Keenable, an independent web index: web search and page fetch. Use it when an agent needs current search results and then needs to read the full content of selected pages as markdown.

1
2
3
4
5
6
7
8
9

from ag2 import Agent
from ag2.config import AnthropicConfig
from ag2.extensions.tools.search import KeenableSearchToolkit

agent = Agent(
    "researcher",
    config=AnthropicConfig(model="claude-sonnet-5"),
    tools=[KeenableSearchToolkit()],
)

Without a key, the toolkit calls Keenable's keyless endpoints, which have lower rate limits than keyed access (see Keenable rate limits for the current numbers). To raise those limits, pass api_key=... or set the KEENABLE_API_KEY environment variable, which is read at call time when api_key is omitted.

Tools#

Shared defaults#

Constructor defaults are applied to the toolkit's default keenable_search tool:

1
2
3
4
5
6

toolkit = KeenableSearchToolkit(
    max_results=5,                 # 1-50, the API default is 10
    site="docs.python.org",        # restrict results to one domain
    published_after="2026-01-01",  # only pages published after this date (YYYY-MM-DD)
    snippet_max_length=2000,       # approximate snippet length in characters (default 1000)
)

snippet_max_length defaults to 1000 characters. Without a cap, Keenable returns several thousand characters of page text per result, so ten results can put tens of thousands of characters into the model's context; use keenable_fetch to read a whole page instead. The value is a hint rather than a hard cap, so a snippet can run slightly over it.

Picking a subset of tools#

Each tool is exposed as a factory method on the toolkit (toolkit.search(), toolkit.fetch()). Call the method to get a ready-to-use tool, then pass only the ones you need to the agent:

1
2
3
4
5
6
7

toolkit = KeenableSearchToolkit()

agent = Agent(
    "researcher",
    config=config,
    tools=[toolkit.search(site="github.com", max_results=10)],
)

Every factory method also accepts name and description overrides for the registered tool. The toolkit.search() defaults (max_results, site, published_after, snippet_max_length) also accept a Variable for deferred context resolution.

Connection settings#

The toolkit opens one httpx.AsyncClient per request. Override how that client connects:

1
2
3
4
5
6

toolkit = KeenableSearchToolkit(
    base_url="https://api.keenable.ai",  # override to route through your own gateway
    timeout=60.0,                        # per-request timeout in seconds
    proxy="http://proxy.internal:3128",
    verify=True,                         # set False only to skip TLS verification
)

A non-2xx response raises httpx.HTTPStatusError with Keenable's own error message, for example when a fetched page does not exist. A rate-limited response also reports its Retry-After value.

Result#

keenable_search returns a KeenableSearchResponse:

snippet holds text extracted from the page, and acquired_at is when Keenable last indexed it.

keenable_fetch returns a KeenableFetchResult:

Request parameters and response fields are documented in the Keenable API docs.