Issuing a Backlog API Key to Drive It from AI Agents

I wired Backlog into Claude Code, Codex, and Hermes Agent by issuing a Backlog API key and letting the agents call the REST API directly. It runs at production level. This post covers the setup that worked, and the two things I want Backlog to improve for AI-agent use: the API key stays visible as plaintext in the browser, and its permissions cannot be scoped down.

From curl to the Backlog CLI 'bee' — Making AI Agent Integration One Step Easier

Last time I wrote about issuing a Backlog API key so AI agents (Claude Code, Codex, Hermes Agent) could call the REST API with curl. This post drops that raw-curl setup and replaces it with bee, the Backlog CLI built by Nulab volunteers. Think of it as the gh of Backlog.

Tavily Alternatives Compared — Cost-Optimizing Search and Extraction APIs for AI Agents

AI agents now run dozens or hundreds of web lookups per task, and Tavily's credit-based pricing has become a real cost driver. This post compares the major Tavily alternatives (Firecrawl, Brave Search, Exa, Serper, Jina Reader, SearXNG) on price, features, migration effort, and industry shifts, then suggests the best combination per workload.