Nbility Blog

Articles about AI Agents, model APIs, developer-tool integration, and Nbility workflows.

Featured

May 25, 2026FeaturedOpenClawAI AgentVPS

OpenClaw Deployment Guide: Run a 24/7 AI Agent on Your Server

Part 5 of the AI Agent Getting Started series: deploy OpenClaw on a VPS, install Node/npm and the CLI, run onboarding, configure a model API, open the Web UI, connect messaging channels, and keep the Agent alive with systemd.

All posts

Jul 21, 2026429Retry

How One 429 Took Down an Agent Chain: Retry, Backoff, and Failover Postmortem

Replay a synthetic 429 retry storm with immediate retries, synchronized exponential backoff, and jitter plus a total budget, then build a safe gateway and agent state machine.

Jul 20, 2026Multi-Model RoutingLLM Router

Multi-Model Routing Is Not Random Switching: Balancing Quality, Cost, and Latency

Use three tasks and a runnable offline replay to design constraint-first LLM routing across quality thresholds, cost, TTFT, health, cache affinity, and fallback.

Jul 19, 2026LLM GatewayAPI Gateway

LLM Gateway vs API Gateway: Follow One Model Request End to End

Trace authentication, model mapping, protocol translation, SSE streaming, retries, token billing, and observability to see what an LLM gateway adds beyond a conventional API gateway.

Jul 19, 2026Prompt CachingOpenAI

Does Prompt Caching Save Money? Replaying Three Bills with the Same 12K Context

Replay OpenAI, Claude, and Gemini cache bills with a 12,000-token prefix and ten requests. Calculate writes, reads, TTL costs, misses, and break-even points from official pricing.

Jul 19, 2026Agent SkillsSKILL.md

Why Agent Skills Took Off: CLAUDE.md vs AGENTS.md vs SKILL.md

Agent Skills are more than prompts saved to files. Learn how CLAUDE.md, AGENTS.md, and SKILL.md differ, then inspect a two-call experiment measuring instruction compliance and token cost.

Jul 19, 2026MCPA2A

MCP vs A2A: One Connects Tools, the Other Connects Agents

MCP and A2A are both agent protocols, but they solve different problems. Compare tool invocation with agent delegation using the same refund workflow, Agent Cards, Tasks, Artifacts, transports, and security boundaries.

Jul 19, 2026MCPModel Context Protocol

What Is MCP? Host, Client, and Server Explained with a Real Tool Call

MCP is not a model API or merely a plugin system. Follow a real order lookup to understand MCP hosts, clients, servers, JSON-RPC, stdio, and Streamable HTTP.

Jul 18, 2026Structured OutputJSON Schema

Structured Outputs and JSON Schema: Reliable JSON from AI Models

Stop malformed or inconsistent AI JSON. Learn the difference between JSON mode and schema-constrained output, with OpenAI, Claude, and Gemini examples plus validation and retry patterns.

Jul 18, 2026OpenAI APIClaude API

OpenAI vs Claude vs Gemini API Formats: A Practical Comparison

Compare OpenAI Responses and Chat Completions, Claude Messages, and Gemini GenerateContent across authentication, messages, tools, streaming, multimodal input, and structured output.

Jul 18, 2026API KeysGitHub

How to Store API Keys Safely and Keep Them Out of GitHub

Learn where API keys belong: environment variables, ignored .env files, secret managers, and CI/CD secrets—plus the correct response when a credential reaches Git history.

Jul 17, 2026Context Length ExceededContext Window

How to Fix Context Length Exceeded: Token Budgets, Trimming, and Compaction

Fix context length exceeded errors by budgeting input and output tokens, trimming messages safely, summarizing history, limiting RAG and tool results, and compacting agent state.

Jul 17, 2026AI APIEmpty Response

AI API Returned Empty Content? Debug content, Reasoning, and Response Formats

A successful AI API call can still produce null or empty text. Learn how to inspect content, tool calls, reasoning, finish reasons, streaming events, and protocol-specific response paths.

Jul 17, 2026WindsurfDevin Desktop

Can Windsurf Use a Third-Party API? Support Limits and Alternatives

Does Windsurf support custom Base URLs, OpenAI-compatible APIs, or BYOK? Learn what Cascade's model selector and proxy settings actually do, plus practical alternatives.

Jul 17, 2026ContinueVS Code

Connect Continue to an OpenAI-Compatible API in VS Code

Configure Continue 2.0 in VS Code with config.yaml, an OpenAI-compatible apiBase, API key, model roles, and fixes for Chat, Edit, Autocomplete, and Responses API issues.

Jul 17, 2026ClineOpenAI-Compatible API

Connect Cline to a Custom API: Models, Permissions, and Cost Control

Configure an OpenAI-compatible Base URL, API key, model, and capabilities in Cline, then control risk and spend with Auto Approve, Checkpoints, .clineignore, and model routing.

Jul 17, 2026Claude CodeCodex

Claude Code vs Codex vs Cursor vs OpenCode: Which Coding Agent Fits You?

Compare Claude Code, Codex, Cursor, and OpenCode across interfaces, models, custom APIs, automation, permissions, cost, and real developer workflows.

Jul 17, 2026Qwen CodeOpenAI-Compatible API

Connect Qwen Code to an OpenAI-Compatible API: Install to First Request

Install Qwen Code, configure an OpenAI-compatible Base URL, API key, and model in settings.json or environment variables, then validate and troubleshoot your first agent task.

Jul 17, 2026Roo CodeOpenAI-Compatible API

Connect Roo Code to a Custom API: Choosing Claude, GPT, or Gemini

Configure an OpenAI-compatible API in Roo Code with the correct Base URL, API key, and model ID, then choose Claude, GPT, or Gemini by coding task.

Jul 16, 2026OpenCodeOpenAI-Compatible API

Connect OpenCode to a Third-Party API: Custom Providers and Multiple Models

Configure a third-party OpenAI-compatible API in OpenCode with /connect and opencode.json, add multiple models, protect credentials, and troubleshoot failures.

Jul 16, 2026CursorOpenAI-Compatible API

Using a Custom OpenAI API in Cursor: Setup, Limits, and Troubleshooting

Configure Cursor BYOK correctly, understand the limits of custom OpenAI-compatible Base URLs, and fix authentication, endpoint, model, and streaming errors.