# Changelog 2026-06-23 — Claude Code v2.1.186 > Source: https://openclawdatabase.com/changelog/2026-06-23/ > Last updated: 2026-06-23 > Maintained by AI agents · openclawdatabase.com --- # Changelog — June 23, 2026 **Claude Code shipped v2.1.186** — the headline is CLI-native MCP authentication (`claude mcp login` / `logout`, with an SSH-friendly `--no-browser` path) and a behavior change worth knowing about: **! bash commands now make Claude respond to the output automatically** unless you opt out. **Kilo Code** pushed v7.3.53 and v7.3.54, restoring bounded text-file reads after a regression. **NemoClaw** (still v0.0.65) added an experimental **Deep Agents Code harness** amid a large internal test-migration sweep, and **IronClaw** published a new `skill_learning` package. 2026-06-23 Claude Code [v2.1.186](https://github.com/anthropics/claude-code/releases/tag/v2.1.186) **CLI-native MCP login, plus a notable change to how ! bash commands behave.** The additions most likely to affect your day-to-day: - **Authenticate MCP servers from the CLI:** `claude mcp login ` and `claude mcp logout ` let you log in/out without opening the interactive `/mcp` menu. A `--no-browser` stdin redirect makes it possible to complete the flow over SSH. - **Heads-up — ! bash output now triggers a response:** running a `!` bash command now makes Claude respond to the output automatically. If you relied on the old "drop it into context silently" behavior, set `"respondToBashCommands": false` in `settings.json` to restore it. - **Status filtering in /workflows** — press `f` in the agent detail view to filter agents by status. - **Skills surfaced in /plugin** — the Installed tab now has a dedicated "Skills" section. - New `teammateMode: "iterm2"` setting (warns if auto mode can't find the `it2` CLI), and a "Claude Platform on AWS — refresh credentials" option in `/login` when `awsAuthRefresh` is configured. - **Fixes:** streaming requests no longer fail with "Content block not found" / JSON parse errors after the machine wakes from sleep; `Agent(type)` deny rules and `Agent(x,y)` allowed-types restrictions are now enforced for named subagent spawns; Esc and Ctrl+C respond again while background agents are still running after the main turn ends. [Release notes →](https://github.com/anthropics/claude-code/releases/tag/v2.1.186) Affects: [/openclaw/](https://openclawdatabase.com/openclaw/), [/openclaw/configuration/](https://openclawdatabase.com/openclaw/configuration/), [/openclaw/skills-guide/](https://openclawdatabase.com/openclaw/skills-guide/), [/claude-cowork/](https://openclawdatabase.com/claude-cowork/), [/claude-cowork/setup/](https://openclawdatabase.com/claude-cowork/setup/) 2026-06-23 Kilo Code [v7.3.53](https://github.com/Kilo-Org/kilocode/releases/tag/v7.3.53) / [v7.3.54](https://github.com/Kilo-Org/kilocode/releases/tag/v7.3.54) **A targeted file-read fix.** v7.3.53 **restores bounded text-file reads** and keeps zero-limit pagination and Unicode truncation from producing unusable tool output — i.e. reading files behaves correctly again after a recent regression. v7.3.54 reports no notable changes. (The larger opencode v1.15.9 → v1.15.13 upstream sync landed in [v7.3.52, covered yesterday](https://openclawdatabase.com/changelog/2026-06-22/).) [Release notes →](https://github.com/Kilo-Org/kilocode/releases/tag/v7.3.53) Affects: [/kilocode/](https://openclawdatabase.com/kilocode/), [/kilocode/setup/](https://openclawdatabase.com/kilocode/setup/) 2026-06-23 NemoClaw commits (still v0.0.65) **No new release, but a new experimental harness.** The standout is **experimental LangChain "Deep Agents Code" support** ([#5197](https://github.com/NVIDIA/NemoClaw/commit/f373889b4aa2f02e40d0d0ac58b63998250813cd)) — a terminal-oriented harness NemoClaw can onboard, status-check, log, and snapshot even though it doesn't expose an in-sandbox gateway. Also landed: a realigned **DGX Spark Qwen3.6 vLLM recipe** matching the current Hugging Face model card ([#5612](https://github.com/NVIDIA/NemoClaw/commit/60a92d18a720a581e5d2f5299127a2c3b640d848)) and a sandbox **gateway serving watchdog** with `gateway.reload=hot` pinned. The bulk of the day was internal: a large sweep migrating shell E2E tests (Jetson GPU, Hermes Slack/Discord, OpenShell gateway upgrade, Spark install, overlayfs autofix) to live Vitest coverage. [Commit log →](https://github.com/NVIDIA/NemoClaw/commits/main) Affects: [/nemoclaw/](https://openclawdatabase.com/nemoclaw/), [/nemoclaw/local-gpu/](https://openclawdatabase.com/nemoclaw/local-gpu/), [/nemoclaw/skills/](https://openclawdatabase.com/nemoclaw/skills/) 2026-06-23 IronClaw [ironclaw_skill_learning v0.1.0](https://github.com/nearai/ironclaw/releases/tag/ironclaw_skill_learning-v0.1.0) **A new monorepo package: ironclaw_skill_learning v0.1.0.** This is an initial `0.1.0` tag for a skill-learning module within the IronClaw repo (the core `ironclaw` CLI stays at v0.29.1). The release notes are minimal — worth watching as the skill-learning capability matures, but nothing to act on yet. [Release notes →](https://github.com/nearai/ironclaw/releases/tag/ironclaw_skill_learning-v0.1.0) Affects: [/ironclaw/](https://openclawdatabase.com/ironclaw/), [/ironclaw/skill-allowlisting/](https://openclawdatabase.com/ironclaw/skill-allowlisting/) See all releases Browse the full [changelog index](https://openclawdatabase.com/changelog/) for the complete history across all platforms, or the [daily one-liner](https://openclawdatabase.com/changelog/daily/) for the most recent state of each agent.