Three places. Choose an agent. Read what changed. Check the evidence behind a skill or MCP server. The rest of the site supports those three jobs.
Registry stats loading…
We separate release-note language from the behavior, policy, price, limit, or compatibility change that matters in practice.
We look for pricing changes, migration risk, reliability regressions, security exposure, and hidden operational limits before recommending an update.
Each report ends with a practical verdict: update, wait, test first, pin a version, change configuration, or ignore the announcement.
Thirteen published items in seven days. Three themes carried it: **Anthropic made Sonnet the new default with five breaking API changes** (Claude Sonnet 5.5 launched at the same $2/$10 list price as Sonnet 5, but thinking:disabled now returns 400, forced tool use returns 400, and computer-use tool declarations on Claude API and Google Cloud reject computer_20251124); **Anthropic turned the Sonnet 4.5 sunset into a 60-day forced migration** (deprecation announced Sept 30, retirement Nov 30, 2026, with the replacement priced 33% lower across input, output, and cache reads); and **the Claude API was unavailable for thirty-eight minutes mid-week** (Claude.ai, Claude Code, Claude Cowork, and the Claude API all elevated errors between 14:00 and 14:59 UTC on Sept 29, with the status page explicitly warning that some messages sent in that window may not have been saved). Plus OpenAI shipped GPT-6.1 Sol with a 50% cache-read cut, an Ultrafast tier on GPT-6 Astra, and computer use added to the Agents API on a single Sept 29 changelog; three Claude Code releases in five days (v2.1.284, v2.1.285, v2.1.287) brought Sonnet 5.5 default, allowedProviders managed setting, a non-streaming retry budget fix, the new Claude Mods plugin behavior layer, and 1M context becoming the implicit default for Opus 4.7+ and Fable on four hosting platforms; two OpenAI Codex drops (v0.158.0 and v0.159.0) added MCP OAuth with pre-registered client secrets, an opt-in bearer token on the direct exec-server WebSocket, an opt-in instant_interrupt live-steering primitive, and .aws directories protected by default under sandbox-writable roots; MCP TypeScript SDK v2.2.0/v1.31.0 shipped a credential-confused-deputy guard via required expectedIssuer and an AuthorizationServerMismatchError; and OpenClaw v2026.8.35 (Oct 2) named GPT-6.1 Sol, stopped durable worker retries from freezing hosts, and made plugin settings survive updates.
We have been writing about agents. This is the map, so you do not have to guess which page does which job.
A short list of agents people actually run. Each dossier says when to pick it, when to skip it, and what it writes. Not a tools dump.
Releases, prices, limits, outages, and compatibility. Each report should say what changed, what could break, and what to do next.
Skills and MCP records, with scan states and gaps. This is not a product catalog and not a safety certificate.
We do not rewrite launch posts. We connect primary sources, reproducible tests, scanner evidence, Registry records, and technical guides to the decision in front of you.
Pricing, policy, security, compatibility, and incident claims start with official documentation, repositories, advisories, or status pages. Community reports are labeled as experience, not proof.
Field reports name the version, environment, method, inputs, and limitations. If we only compared documentation, we say so.
The skills and MCP Registry remains free. Its source links, scan states, provenance, and explicit evidence gaps help test whether an ecosystem claim survives contact with real artifacts.
Durable technical guides and inspectable workflow manifests turn a news event into implementation work you can adapt and validate in your own environment.
Anthropic's coding agent CLI. Reads project instructions, calls tools, and stays human-gated per session.
OpenAI's coding agent CLI. Tool-using and session-gated. Not the hosted Agents API.
Self-hosted agent gateway with channels, cron, and write-tools. Failure has an operator.
A landing page that says "AI agent" is not a listing. The directory covers coding, ops, voice, research, and go-to-market runtimes. Autonomy is a tested facet, not a slogan.
Weekday impact reports, a weekly Before You Update brief, and tested comparisons when the evidence supports them. No quota filler.
The change log is the news. Agent dossiers are the catalog.
Open the change log