AI agents for
autonomous work.

Default when the agent ships inside an existing Next.js or Node app and the team wants the runtime co-located with the rest of the product.

import { streamText } from "ai";
import { openai } from "@ai-sdk/openai";

const result = await streamText({
  model: openai("gpt-4.1"),
  prompt: ticket.body,
  tools: { searchKb, jiraCreate },
});

First-class agent loop with tools, file editing, and bash. Picked when the work is multi-step reasoning, code, or long-context document workflows.

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic();
const stream = client.messages.stream({
  model: "claude-opus-4-7",
  tools: [searchKb, jiraCreate],
  messages: [{ role: "user", content: ticket.body }],
});

Responses API and function calling. The right pick when the team is already on the OpenAI stack or needs the managed runtime + web search out of the box.

from openai import OpenAI

resp = OpenAI().responses.create(
    model="gpt-4.1",
    instructions="You triage tickets for Acme.",
    input=ticket.body,
    tools=[search_kb, jira_create],
)

Model-agnostic Agent Development Kit centred on Gemini and Vertex. Picked when the data, IAM, and observability already live in GCP.

from google.adk.agents import Agent
from google.adk.runners import Runner

agent = Agent(
  name="triage",
  model="gemini-2.5-pro",
  tools=[search_kb, jira_create],
)
Runner(agent).run(ticket.body)

Standard install · lint · eval · deploy workflow. Every promotion runs the regression suite and posts the delta on the PR before a human merges.

# .github/workflows/deploy.yml
jobs:
  ship:
    runs-on: ubuntu-latest
    steps:
      - run: pnpm install
      - run: pnpm eval --gate 0.92
      - run: pnpm deploy --env prod

Agentic steps inside CI. Triages failed eval runs, drafts postmortems, and grades PR diffs against the agent's expected behaviour before they reach prod.

# .github/workflows/eval-grade.yml
jobs:
  grade:
    steps:
      - uses: anthropics/claude-code-action@v1
        with:
          prompt: "Grade eval delta vs main."

Same shape as the GitHub workflow, on the runner pool your team already operates. Picked when GitLab is the company default.

# .gitlab-ci.yml
ship:
  image: turion/agentic-runner:1
  script:
    - npm ci
    - npm run eval -- --gate 0.92
    - npm run deploy -- --env prod