Agent runtimes, explored
Comparisons and guides to help you choose how to build, deploy, and run your agents.
Other content
Java October 30, 2026 AI agent frameworks in Java: Spring AI, LangChain4j, Embabel (2026) A Java agent framework comparison for 2026: the Spring AI 2.0 release and Spring AI 1.1, LangChain4j release 2026 (1.23.0) with agentic human in the loop, the Embabel agent framework release, Google ADK Java, JetBrains Koog and LangGraph4j checkpoints, with tool calling, chat memory, and the Temporal Spring AI integration. LangGraph October 30, 2026 LangGraph in production (2026) LangGraph durable execution, interrupts, and persistence in production: what the Postgres checkpointer stores, LangGraph interrupt human in the loop at scale, resumable stream reconnect on the Agent Server, LangSmith Deployment (formerly LangGraph Platform) pricing and self-hosted options, queue workers, and scaling. Mastra October 30, 2026 Mastra for AI agents (2026) Mastra for AI agents in 2026: what Mastra does since the Mastra 1.0 release, Mastra workflows suspend resume with Postgres storage, durable agents and durable execution, tool approval and human in the loop after a restart, observational memory and the model router, MCP and OpenAPI tools, and Mastra Cloud pricing on the Mastra platform. Code execution October 30, 2026 Running agent code in Blaxel (2026) Run agent-generated code in a Blaxel sandbox from a Cadenya agent: a bare run_code tool that your own worker answers with Blaxel's SDK, so you expose no endpoint and the Blaxel key stays on your servers. With the config files, the TypeScript worker, a run on Blaxel, an approval before code runs, and an HTTP tool set for teams that host an endpoint. Code execution October 30, 2026 Running agent code in Daytona (2026) Run agent-generated code in a Daytona sandbox from a Cadenya agent: a bare run_code tool that your own worker answers with Daytona's TypeScript SDK, so you expose no endpoint and the Daytona key stays on your servers. With the config files, the TypeScript worker, a run on Daytona, an approval before code runs, and an HTTP tool set for teams that host an endpoint. Code execution October 30, 2026 Running agent code in E2B (2026) Run agent-generated code in an E2B sandbox from a Cadenya agent: a bare run_code tool that your own worker answers with E2B's Code Interpreter SDK, so you expose no endpoint and the E2B key stays on your servers. With the config files, the TypeScript worker, a run on E2B, an approval before code runs, and an HTTP tool set for teams that host an endpoint. Go October 15, 2026 AI agent frameworks in Go: Eino, Genkit, LangChainGo, trpc-agent-go (2026) A Go agent framework comparison: the Genkit Go 1.0 release and its interrupts, CloudWeGo Eino ADK interrupt, checkpoint and resume, LangChainGo maintenance status and latest release, trpc-agent-go graph interrupt and checkpoint, and multi-provider support for Anthropic, OpenAI and Gemini. OAuth October 15, 2026 AI agents acting on behalf of the signed-in user (2026) How AI agents act on behalf of the signed-in user: OAuth 2.0 token exchange (RFC 8693) with on-behalf-of delegation and the act claim as an audit trail, the MCP enterprise-managed authorization extension with ID-JAG, and token exchange support in Keycloak, Spring Authorization Server, Okta, Auth0 and Microsoft Entra. Ruby on Rails October 15, 2026 AI agents in Ruby on Rails: RubyLLM, ActiveAgent, Raix (2026) AI agents in Ruby on Rails: the RubyLLM gem for multi-provider Anthropic, OpenAI and Gemini, acts_as_chat tools, streaming and persistence, requires_approval and compaction in RubyLLM 2.0, the ActiveAgent gem, Raix, and a hosted agent loop with a Ruby SDK. Claude Agent SDK October 15, 2026 Claude Agent SDK in production (2026) Claude Agent SDK hosting in production: the subprocess and containers you run, sessions resume and persistence, multi-tenant isolation, canUseTool permissions, compaction, Bedrock and Vertex, and a hosted agent loop with nothing to run. Cloudflare Agents SDK October 15, 2026 Cloudflare Agents SDK: AIChatAgent, resumable streaming and tool approval Cloudflare Agents SDK resumable streaming, AIChatAgent human in the loop tool approval with needsApproval, context compaction and maxPersistedMessages, useAgentChat in Vue, the MCP client, and Mastra vs LangGraph.js. Chat UI October 15, 2026 CopilotKit vs assistant-ui vs AG-UI for AI agent chat UIs (2026) CopilotKit vs assistant-ui vs AG-UI protocol for AI agent chat UIs: frontend tools (useFrontendTool, makeAssistantTool), human in the loop, Vue, Angular and vanilla JavaScript without React, the AG-UI protocol Go SDK, and thread persistence. Django October 15, 2026 Durable AI agents in Django with Celery (2026) Durable AI agents in Django: Celery human-in-the-loop approvals, the Django 6.0 tasks framework, the DBOS Django integration on Postgres, the Temporal Python SDK with signals for durable execution, and SSE streaming under ASGI. Go October 15, 2026 Durable AI agents in Go: Temporal Go SDK, Restate, DBOS and Hatchet (2026) Durable AI agents in Go: the Temporal Go SDK, the Restate Go SDK, DBOS Transact Go durable workflows on Postgres, and Hatchet Go SDK durable tasks, compared on the latest release, human-in-the-loop signals and awakeables, what you self-host, license, and a hosted agent loop with a Go SDK. Durable execution October 15, 2026 Durable AI agents in Rails: Temporal Ruby SDK, Solid Queue and Sidekiq (2026) Durable AI agents in Rails: Temporal Ruby SDK general availability, Rails 8.1 ActiveJob::Continuable job continuations, Solid Queue vs GoodJob vs Sidekiq for Postgres background jobs, acidic_job, ChronoForge, and Restate and Hatchet Ruby SDKs. Google ADK October 15, 2026 Google ADK agents in Python (2026) Google ADK agents in Python: the session service (DatabaseSessionService or VertexAiSessionService), tool confirmation for human-in-the-loop, resumability, OpenAPIToolset, LiteLLM for Claude, context compaction, and deployment to Agent Engine or Cloud Run. Google ADK October 15, 2026 Google ADK for Go agents (2026) Google ADK for Go (adk-go) in 2026: the release version and module path, function tools and the OpenAPI toolset, human-in-the-loop tool confirmation and long-running function tools, the session service database, Anthropic Claude and OpenAI model support, and the Temporal integration. Durable execution October 15, 2026 Hatchet for AI agents (2026) Hatchet for AI agents: Hatchet durable tasks, WaitForEvent for human in the loop in the Go SDK and the Ruby SDK (hatchet-sdk gem), SDK languages Python, TypeScript, Go and Ruby, self-host on Postgres with RabbitMQ optional under the MIT license, and Hatchet Cloud pricing per task runs. Inngest October 15, 2026 Inngest AgentKit: durable agents, streaming and Go support (2026) Inngest AgentKit for AI agents: durable agents, step.waitForEvent human in the loop, realtime streaming with useAgent, step.ai.infer and step.ai.wrap, Inngest Go SDK support, pricing and free tier, and self-hosting with Postgres. Inngest October 15, 2026 Inngest for Python AI agents (2026) Inngest for Python AI agents: the Inngest Python SDK release and Python version, durable execution in Python with step.run, step.wait_for_event for approvals, Inngest AgentKit and Python, SDK languages, the SSPL self-hosting license, and pricing in executions. LangGraph October 15, 2026 LangGraph.js human-in-the-loop in TypeScript (2026) LangGraph.js human-in-the-loop in TypeScript: interrupt and Command resume, the PostgresSaver checkpointer, durability mode, createAgent middleware, useStream, and how Mastra, Inngest AgentKit, and the Cloudflare Agents SDK compare. OpenAI Agents SDK October 15, 2026 OpenAI Agents SDK sessions: SQLAlchemy, Redis and hosting (2026) OpenAI Agents SDK sessions in Python and TypeScript: human-in-the-loop with RunState serialize and resume, SQLAlchemySession and Redis session persistence, SQLiteSession, OpenAIConversationsSession, OpenAIResponsesCompactionSession, and what you host in production. Pydantic AI October 15, 2026 Pydantic AI agents in production (2026) Running Pydantic AI agents in production: Pydantic AI durable execution with Temporal, DBOS and Prefect, deferred tools with requires_approval and DeferredToolRequests, the OpenAPI toolset question, history processors and compaction, and the v2 release and version policy. Slack October 15, 2026 Slack Bolt for Go and Ruby (2026) Is there a Slack Bolt for Go or Ruby? Slack Bolt SDK official languages are JavaScript, Python and Java, so Go apps use slack-go/slack and Ruby apps use slack-ruby-client, with socket mode, assistant threads and chat.startStream, plus a hosted agent loop with Go and Ruby SDKs. Trigger.dev October 15, 2026 Trigger.dev for AI agents (2026) Trigger.dev for AI agents in 2026: chat.agent with the AI SDK useChat, realtime streams, Trigger.dev pricing, v4 self-hosting with Docker or Kubernetes (Postgres, Redis, ClickHouse), version locking and atomic deploys on Vercel, compared with Inngest, Temporal and Restate. TypeScript October 15, 2026 TypeScript AI agent frameworks compared: Mastra vs LangGraph.js vs AI SDK vs OpenAI Agents JS (2026) TypeScript AI agent frameworks compared: Mastra vs LangGraph.js vs Vercel AI SDK vs OpenAI Agents JS on workflows suspend and resume, tool approval (requireApproval, toolApproval, needsApproval), PostgresStore and the PostgresSaver checkpointer, the Mastra model router, observational memory, and the aisdk adapter. Agent memory October 8, 2026 Agent memory and context windows (2026) How AI agents keep conversation state in 2026: short-term memory vs long-term memory, context compaction when the context window fills, LangGraph checkpointers and SummarizationMiddleware, OpenAI Responses API compaction, Anthropic context management, Mastra, Pydantic AI, and the AI SDK. Vercel AI SDK October 8, 2026 Durable agents with the Vercel AI SDK (2026) How to make a Vercel AI SDK agent durable: AI SDK tools (tool calling with tool(), multi-step with stopWhen, tool approval), resumable streams with useChat resume and persistence, tool approval with toolApproval and needsApproval, human in the loop waits, WorkflowAgent, and the Temporal, Inngest, DBOS and Restate integrations. Durable execution October 8, 2026 Durable execution for AI agents (2026) Durable execution and durable workflows for AI agents in 2026: Temporal, Restate, Inngest, DBOS, Vercel Workflow, LangGraph checkpointers, and Mastra suspend and resume compared on surviving crashes and deploys, human-in-the-loop waits, wait for event, worker versioning, what you run, and pricing. Human in the loop October 8, 2026 Human-in-the-loop approvals for AI agents (2026) Human-in-the-loop approvals for AI agents in 2026: tool approval in LangGraph interrupts (interrupt() and HumanInTheLoopMiddleware), OpenAI Agents SDK needs_approval and RunState, the Vercel AI SDK, Temporal signals, Inngest waitForEvent, and Claude Managed Agents, with where the paused state lives and what deny does. OpenAPI October 8, 2026 OpenAPI spec to AI agent tools or an MCP server (2026) Turning an OpenAPI spec into AI agent tools or an MCP server in 2026: FastMCP from_openapi, Stainless and Speakeasy MCP server generators, AWS AgentCore Gateway, the LangChain OpenAPI toolkit, and how to keep tools in sync, fit a big API in the context window, and call the API as each user with MCP OAuth on a remote MCP server. Retries October 8, 2026 Retries and idempotency for AI agent tool calls (2026) Retries and idempotency for AI agent tool calls: retry policy defaults, at least once vs exactly once, idempotency keys for external side effects, and timeouts in Temporal, Restate, Inngest, DBOS, Trigger.dev, and plain queues. Code execution October 8, 2026 Sandboxes for running agent-generated code (2026) Code sandbox options for running LLM-generated code from AI agents in 2026: E2B, Daytona, Modal Sandboxes, Vercel Sandbox, and Cloudflare Sandbox SDK compared on isolated code execution (Firecracker, gVisor, containers), cold start, E2B pricing, Daytona pricing and Modal sandboxes pricing, persistence, SDKs, and self-hosted sandbox options, plus the sandbox support in the OpenAI Agents SDK and the Claude Agent SDK. Self-hosted October 8, 2026 Self-hosted or managed AI agent infrastructure (2026) Self-hosted or managed AI agent infrastructure in 2026: what you run to self-host Temporal, Restate, Inngest, DBOS, LangGraph, Trigger.dev and Hatchet (services, Postgres, Redis, single binary), the server license, the operational burden, and the managed option's starting price. Slack October 8, 2026 Slack AI agents with human in the loop approvals (2026) How to build a Slack AI agent in 2026: Slack Bolt and assistant threads, socket mode vs Events API, acknowledge events within 3 seconds, human in the loop approval buttons, thread persistence, act on behalf of user, and where LangGraph, the OpenAI Agents SDK, Chat SDK, Temporal, Inngest and the Cursor SDK fit. Model routing October 8, 2026 Switching model providers for AI agents (2026) How to keep an AI agent provider agnostic in 2026: LiteLLM vs OpenRouter vs Vercel AI Gateway, the AI SDK provider registry, Portkey, and Cloudflare AI Gateway compared on model routing, provider routing, fallbacks, BYOK, markup and fees, OpenAI compatible and Anthropic Messages endpoints, and self-hosting. AWS AgentCore September 28, 2026 AWS Bedrock AgentCore alternatives in 2026 Amazon Bedrock AgentCore alternatives for 2026. AgentCore hosts the agent you write, with any model, on AWS. The alternatives differ on who writes the agent loop, how your API becomes tools, how approvals work, and what you pay for. Claude Managed Agents September 28, 2026 Claude Managed Agents alternatives in 2026 Claude Managed Agents alternatives for 2026. What it does well, where it stops (Claude only, no OpenAPI import), and which alternative answers each, with pricing. Durable execution September 28, 2026 DBOS vs Restate for AI agents (2026) DBOS vs Restate for durable execution and human-in-the-loop approvals in AI agents: a library on your Postgres vs a single self-hosted binary, waits, versioning, MIT vs BSL, pricing, and a hosted option. Durable execution September 28, 2026 DBOS vs Temporal for AI agents (2026) DBOS vs Temporal for durable execution and human-in-the-loop approvals in AI agents: a library on your Postgres vs a server, send() and recv() vs Signals, versioning, pricing, and a hosted option. Hosted agent runtimes September 28, 2026 Hosted agent runtimes in 2026 Five hosted agent runtimes compared for 2026: pricing, which models you can use, who writes the agent loop, human-in-the-loop approvals, streaming, and your own API as tools. Durable execution September 28, 2026 Inngest vs Restate for AI agents (2026) Inngest vs Restate for durable execution and human-in-the-loop approvals in AI agents: two single binaries compared on waiting primitives, versioning, self-hosting licenses (SSPL vs BSL), SDKs, pricing, and a hosted option. Durable execution September 28, 2026 Inngest vs Temporal for AI agents (2026) Inngest vs Temporal for durable execution and human-in-the-loop approvals in AI agents: step.waitForEvent vs Signals, what you self-host, versioning, pricing, license, and a hosted option. LangSmith Deployment September 28, 2026 LangSmith Deployment pricing and alternatives in 2026 What LangSmith Deployment, formerly LangGraph Platform, costs in 2026: seats, traces, and metered runtime, with a month of a Dedicated Small deployment worked out, and the LangGraph deployment options: Cloud, hybrid, standalone server, and langgraph dev. Then the alternatives, from AgentCore to a hosted agent loop. Durable execution September 28, 2026 Restate vs Temporal for AI agents (2026) Restate vs Temporal for durable execution and human-in-the-loop approvals in AI agents: what you self-host, the waiting primitive, versioning, pricing, license, and a hosted option. Temporal September 28, 2026 Temporal Cloud vs self-hosted Temporal: the real cost in 2026 Temporal Cloud pricing for an AI agent in 2026, worked out for 2,000 runs a month, against self-hosted Temporal's requirements and a hosted agent loop with no workers to run.
Grow wherever AI goes next.
Start shipping agents that are equipped to evolve.
