The short list of SDKs we use in production, picked because they survive contact with real workloads. Every Turion engagement settles on exactly one of these per agent, then ships through one of the CI/CD pipelines listed below.
TypeScript-first SDK for streaming, structured outputs, and tool use. The default when the integration ships inside an existing Next.js or Node app and the team wants the agent on the same edge runtime as the rest of the product.
First-class agent loop with tool use, file editing, and bash access. Picked when the work involves multi-step reasoning, code, or long-context document workflows where Claude is the strongest model.
Responses API and function-calling primitives in TypeScript and Python. The right pick when the agent needs the latest GPT models, web search, or the OpenAI-managed runtime — and the team is already on the OpenAI stack.
Model-agnostic Agent Development Kit centred on Gemini and the Vertex AI stack. The default when the data, IAM, and observability already live in GCP.
Standard CI/CD: install, lint, run the eval suite, deploy. Every Turion engagement ships through a workflow your team owns and can read on day one.
Agentic workflow steps that run an agent inside CI to triage failures, draft PR descriptions, or grade an eval delta against a golden set before promotion.
GitLab Duo Agent Platform for orchestrating multi-step automation flows from `.gitlab-ci.yml`. The right pick when the rest of the company already runs on GitLab and the agent needs the same runner pool.
Pair with one of our solutions architects. Two weeks from kickoff to a deployed, evaluated, observable agent in your stack.