Skip to content

ebarti/agent-runtime-kit

Repository files navigation

agent-runtime-kit

agent-runtime-kit is a small Python runtime layer for agent SDKs. It gives applications one typed async API for dispatching an agentic task through Claude Agent SDK, OpenAI Codex SDK, or Google Antigravity SDK while keeping provider capabilities visible.

Tags: agent-sdk agent-runtime coding-agents claude-code openai-codex google-antigravity mcp typed-python async-python developer-tools

About

agent-runtime-kit is for Python developers who want to run coding-agent tasks through vendor runtimes without rewriting their application around each SDK. It normalizes the runtime boundary: task inputs, capability checks, event streams, tool audits, package/readiness diagnostics, and typed results.

The library keeps vendor differences visible. Claude, Codex, and Antigravity still expose different capabilities, permission models, setup requirements, and unsupported fields. agent-runtime-kit gives those differences a consistent shape instead of hiding them behind a lowest-common-denominator wrapper.

The package is intentionally not a router, benchmark harness, queue, hosted service, or full agent framework. It is the reusable layer underneath those systems: task models, runtime capabilities, event sinks, package/readiness diagnostics, and adapters.

Install

If you want all first-party runtimes available, install the all extra:

pip install "agent-runtime-kit[all]"

Install the vendor-SDK-free core when you only need the public models, fake runtime, registry, diagnostics types, or you plan to add provider SDKs later:

pip install agent-runtime-kit

Install a single provider extra when your application only dispatches through one vendor runtime:

pip install "agent-runtime-kit[claude]"
pip install "agent-runtime-kit[codex]"
pip install "agent-runtime-kit[antigravity]"

Provider extras are a packaging boundary, not a separate API. They keep the core importable without vendor SDKs, avoid forcing every user to install every CLI binary or compiled runtime wheel, and contain dependency drift when one fast-moving vendor SDK changes independently of the others. Missing adapters raise typed setup errors that point to the matching extra.

import asyncio

from agent_runtime_kit import AgentTask, FakeAgentRuntime


async def main() -> None:
    runtime = FakeAgentRuntime(output="done")
    result = await runtime.run(AgentTask(goal="Summarize this repository"))
    print(result.output)


asyncio.run(main())

The core package has no Claude, Codex, or Antigravity dependency. Vendor SDKs are added through optional extras.

Real Providers

AgentKit is the keyword-native hub: it registers the built-in runtimes, caches them per kind, and turns Python types into structured output.

import asyncio
from dataclasses import dataclass

from agent_runtime_kit import AgentKit, ReadinessStatus


@dataclass
class RepoSummary:
    name: str
    languages: list[str]


async def main() -> None:
    async with AgentKit() as kit:
        readiness = await kit.readiness_for("claude")
        if readiness.status is ReadinessStatus.NOT_READY:
            raise RuntimeError(readiness.message)
        result = await kit.run(
            "claude",
            goal="Summarize this repository",
            permissions="strict",
            output_type=RepoSummary,
        )
        print(result.parsed.languages if result.parsed else result.error)


asyncio.run(main())

Adapters also work standalone when you need vendor-specific configuration:

import asyncio

from agent_runtime_kit import AgentTask, ReadinessStatus, check_readiness
from agent_runtime_kit.adapters import ClaudeAgentRuntime


async def main() -> None:
    runtime = ClaudeAgentRuntime(default_model="claude-sonnet-4-6")
    readiness = await check_readiness(runtime)
    if readiness.status is ReadinessStatus.NOT_READY:
        raise RuntimeError(readiness.message)
    result = await runtime.run(AgentTask(goal="Summarize this repository"))
    print(result.output)


asyncio.run(main())

Runtime Fields

AgentTask supports goal, system prompt, model, reasoning effort, working directory, permission profile, MCP stdio servers, session/resume handles, output schema, budget, metadata, and an async event sink. (model and reasoning_effort are first-class fields; the legacy metadata["model"] / metadata["reasoning_effort"] aliases still work.) Where a runtime cannot honor a field (for example only Claude maps budget_usd; Codex and Antigravity reject it with a typed UnsupportedTaskInputError) the adapter raises rather than silently dropping it.

Model selection follows one explicit precedence chain: AgentTask.model, then legacy metadata["model"], then an adapter's default_model= constructor override, then the provider's native configuration/default. Adapters omit the vendor model option at the final step rather than pinning a library-owned model. Results record metadata["model_source"]; metadata["model"] appears only when the kit knows the selected value.

Call validate_task(runtime, task) (or kit.validate_task("codex", task)) to inspect every statically detectable incompatibility before dispatch. The returned TaskSupportReport is side-effect-free and lists source fields such as budget_usd and permissions.network; run() still fails closed on the first issue. Third-party runtimes can opt into provider-specific checks with the TaskSupportProvider protocol, while older AgentRuntime implementations continue to work through capability-based fallback checks.

availability() is deliberately synchronous, side-effect-free, and package-only. Use await check_readiness(runtime) (or kit.readiness_for(...)) for an explicit, bounded credential/setup probe. READY_TO_ATTEMPT means setup was positively detected, not that a future provider call is guaranteed to succeed; INDETERMINATE lets callers decide whether to attempt provider-chain or local login authentication. Probe failures and timeouts never include credential values.

AgentResult returns output, finish reason (see FinishReason), locally validated structured output, usage, cost, session id, tool-call audits, and provider metadata. is_success is true only for a natural, error-free completion. Unknown usage and cost fields are None; reported zero remains distinct. parsed_output_available distinguishes an absent payload from valid JSON null (both use parsed_output=None). artifacts is a reserved field: no built-in runtime populates it yet, so it is always an empty tuple today.

Docs

Packages

 
 
 

Contributors

Languages