Deuz A.I

Deuz is an AI research and technology company building open intelligence infrastructure for models, applications, and autonomous agents.

[Deuz A.I @ai_deuz]

🚀 Deuz SDK v1.8.0 is live!

New: Autonomous Agent Runtime, persistent workspace, CodeAct, Planner–Executor–Verifier, background runs, browser control, model routing, Azure & Bedrock support, and deterministic agent testing.

Score: 69.6 → 74.0
Rank: #14 → #9

[Deuz A.I @ai_deuz]

Deuz SDK v1.7.1 is live.

This patch hardens the v1.7 runtime for production: strict chat isolation, preserved history, reliable resumed tools, buffered result parity, and complete memory outputs.

No new APIs. No breaking changes. Just a more reliable runtime.

[Deuz A.I @ai_deuz]

Deuz SDK v1.7.0 is live 🚀

Your chatbot can now remember users, stream live USD costs, enforce budgets, survive refreshes and crashes, verify signed tool approvals, and fail over across providers—all without a hosted gateway.

https://t.co/7u4Ptu5zoq

[Deuz A.I @ai_deuz]

New Update — Deuz SDK v1.6.0 is live.

This release introduces the Observable Runtime: local-first visibility into model calls, agent steps, tools, retries, approvals, checkpoints, compaction and sub-agents.

No hosted service. No OTel dependency. Nothing captured by default.

[Umutcan Edizaslan @UEdizaslan]

Deuz SDK v1.5.1 is live.

No flashy features this time—just the hard work that makes everything else possible: stability, compatibility, safer streaming, and a stronger foundation.

Now that the core is solid, I’ve officially started building v1.6.0.

This is where Deuz begins to level up.

https://t.co/bfeBB7TbDh

[Deuz A.I @ai_deuz]

Deuz SDK v1.5.0 is live 🚀

Durable sessions are here:
agents can now checkpoint, pause for approvals, resume safely, and recover without starting from scratch.
Sub-agent approvals now work too, with signed approval tokens for safer production flows.

npm install @deuz-sdk/core

[Deuz A.I @ai_deuz]

Deuz-SDK v1.4.0 is out.

We’re moving beyond chat wrappers.

This release adds Agent Core: sub-agents, automatic compaction, budget-aware stopping, step-level control, and loop hooks.

https://t.co/35Z9OGfGiw

[Umutcan Edizaslan @UEdizaslan]

Deuz-SDK v1.3.0 is here.

From multi-provider AI calls to real agentic app infrastructure:

tool approvals, streaming structured output, MCP resources/prompts, elicitation, and React hooks.

309 tests green.

Less lock-in. More control. Built for production AI apps.

https://t.co/z3YWSXL20b

https://t.co/sVuMTi7fNV

[Deuz A.I @ai_deuz]

Deuz-SDK v1.1.1 is out.

This update improves Claude 5 effort wiring, adds xhigh/max reasoning levels, fixes Gemini thinking behavior, updates the 2026 model catalog, and improves pricing + usage tracking.
Provider-agnostic AI infra keeps getting sharper.
https://t.co/oJ8bN29Iix

[Deuz A.I @ai_deuz]

We’re building Deuz-SDK to rival AI SDK.

A TypeScript-first, provider-agnostic layer for OpenAI, Anthropic, Gemini, xAI, Vertex and more.

Edge-first. Stream-native. Built for tools, memory, RAG, MCP and production AI apps.

Less lock-in. More control.

https://t.co/7u4Ptu5zoq

[Deuz A.I @ai_deuz]

Try our autonomous agent (beta).

https://t.co/bhiU0svmvL