HybridClaw v0.12.5

Release Date: April 13, 2026

The operator-workflow and reliability release — versioned admin agent files, keyboard-driven TUI approvals, safer config recovery, and more predictable email, WhatsApp, and retry-aware transport behavior.

---

Highlights

• Versioned admin agent-file editor — Added /admin/agents so operators can edit each registered agent's allowlisted bootstrap markdown files, review saved revisions, and restore earlier versions directly from the browser.
• Keyboard-driven TUI approvals — hybridclaw tui now opens pending approvals in an interactive picker with arrow-key navigation, number-key quick select, non-interactive fallback prompts, and a resumable exit summary.
• Safer runtime-config recovery — Interactive onboarding can now restore the last known-good saved config snapshot or roll back to the newest tracked revision when config.json becomes invalid.

Changed

• Destructive admin confirmations — Browser-based operator flows now use explicit confirmation dialogs before restarts, deletes, and similar destructive actions.
• Operator docs and release surfaces — README, changelog, and docs landing pages now reflect the current admin-agent, TUI approval, config recovery, and transport-reliability workflows.

Fixed

• Approval replay and exit summaries — Replayed approval prompts now keep their cached details more reliably, web /approve flows preserve pending metadata, and TUI exit summaries consistently show remote usage/tool/file totals or an explicit unavailable state.
• Retry-aware transport backoff — Channel transports now honor service-provided Retry-After delays and reject invalid retry values early instead of silently retrying with bad timing.
• Email first-sync cursor handling — The built-in email transport now seeds a missing mailbox cursor from the current mailbox head so old inbox mail is not replayed as new traffic on first startup.
• WhatsApp startup reliability — The built-in WhatsApp transport now disables Baileys init queries that can trigger intermittent 400/bad-request failures during startup and pairing.

Contributors

Core

• @furukama — admin agent-file editor, TUI approval workflow, runtime-config recovery, transport reliability fixes, and release documentation
• @maxnoller — destructive admin confirmation dialogs and console workflow polish

All Contributors

@furukama, @maxnoller

---

Full Changelog: v0.12.4...v0.12.5

HybridClaw v0.12.4

Release Date: April 12, 2026

The Slack, scheduler, and operator-workflow release — built-in Slack transport, immediate one-shot scheduler jobs, per-agent skill scoping, accessible admin dialogs and toast notifications, and clearer approval presentation across chat channels and gateway surfaces.

---

Highlights

• Slack transport and setup workflow — Added a built-in Slack Socket Mode channel with hybridclaw auth login slack, DM/channel policy controls, thread-aware session routing, approval buttons, file handling, and a dedicated operator setup guide.
• Immediate one-shot scheduler jobs — Added config-backed one_shot jobs that run immediately, retry up to maxRetries, preserve review state, and surface richer delivery output across the gateway and admin scheduler UI.
• Per-agent skill scoping and admin feedback UX — Added per-agent skill allowlists while replacing inline console banners with accessible dialog and toast primitives, tightening scheduler/jobs feedback and improving the mobile admin shell.

Changed

• Admin console interactions — The embedded admin console now uses dialog/toast feedback instead of inline banners, and the mobile topbar/sidebar flow is cleaner for narrow-screen operator work.
• Per-agent skill filtering — Agent skills settings now narrow the globally enabled skill set, while omitting skills preserves the existing global scope for backward compatibility.
• Approval presentation across channels — Gateway approval copy and channel actions are more consistent across Discord, Slack, and gateway-managed approval surfaces.

Fixed

• Approval id preservation — Resumed approval flows preserve approval ids correctly so existing approval tracking stays stable across gateway and chat surfaces.

Contributors

Core

• @furukama — Slack transport, one-shot scheduler jobs, per-agent skill filtering, approval-flow polish, docs updates, and release preparation
• @maxnoller — Accessible dialog/toast components, scheduler/admin feedback polish, and mobile topbar refinement

All Contributors

@furukama, @maxnoller

---

Full Changelog: v0.12.3...v0.12.4

HybridClaw v0.12.3

Release Date: April 12, 2026

The transport, memory, and operator-workflow release — built-in Telegram support, first-class memory inspection, native LOCOMO evals, the bundled GBrain plugin and manim-video skill, refreshed admin navigation, and safer runtime polish across model routing, media staging, and config hot reloads.

---

Highlights

• Telegram transport and inbound media flow — Added a built-in Telegram Bot API channel with BotFather token setup, DM/group policy controls, admin Channels support, inbound media handling, and canonical telegram:<chatId> send targets.
• Memory inspection and native LOCOMO evals — Added /memory inspect, /memory query, and hybridclaw gateway memory inspect diagnostics plus managed and local locomo eval flows with official dataset setup, QA and retrieval modes, detached logs, and retrieval sweeps.
• Bundled GBrain plugin and manim-video skill — Shipped the bundled gbrain plugin for prompt-time recall and gbrain_* tools, alongside a repo-bundled manim-video skill with setup helpers, reference packs, and render guidance for scripted explainer videos.

Changed

• Admin console navigation and email workflow — Reworked the embedded console around a structured sidebar taxonomy, richer channel and email surfaces, and a built-in mailbox UI for browsing folders, inspecting metadata, and composing or replying without leaving the runtime.
• Provider-aware model catalogs and routing — Updated /model list and selector/status surfaces to use provider-scoped catalogs for Codex, OpenRouter, Mistral, and Hugging Face, with explicit openai-codex/... ids and more consistent discovered-model metadata.
• Shared runtime media staging and config reloads — Moved inbound media downloads for email, Telegram, WhatsApp, and Microsoft Teams into the shared uploaded-media-cache directory and taught the gateway to restart Telegram automatically when telegram.* config changes land.

Fixed

• TUI sandbox preflight correctness — hybridclaw tui now respects the sandbox mode reported by a reachable gateway instead of forcing unnecessary rebuild/preflight checks.
• HybridAI auxiliary model prefix handling — Auxiliary routing now strips leading provider prefixes correctly so already-namespaced HybridAI model names do not fail at request time.
• GBrain discovery resilience — The bundled GBrain plugin now times out cleanly when gbrain --tools-json hangs and reports parse failures with stdout/stderr previews to make discovery failures actionable.

Contributors

Core

• @furukama — Telegram transport, LOCOMO eval workflows, GBrain plugin + manim-video skill, built-in memory inspection, admin email surfaces, provider/model routing updates, and runtime robustness fixes
• @maxnoller — Admin console sidebar taxonomy, navigation refinement, and supporting console UX polish

All Contributors

@furukama, @maxnoller

---

Full Changelog: v0.12.2...v0.12.3

HybridClaw v0.12.2

Release Date: April 10, 2026

The memory-provider and operator UX release — bundled Honcho and MemPalace memory plugins, plugin dependency install and health checks, sortable admin tooling with managed gateway restart UX, Discord concierge buttons, and safer runtime defaults.

---

Highlights

• Honcho memory plugin — Added the bundled honcho-memory plugin so HybridClaw can mirror sessions into Honcho, inject prompt-time recall and Honcho tools into later turns, and promote native user-profile saves into Honcho conclusions while built-in memory stays active.
• MemPalace memory plugin — Added the bundled mempalace-memory plugin with /mempalace ... controls, transcript mining, native-memory mirroring, and prompt-time recall that can route through CLI helpers or a configured mempalace MCP server.
• Plugin dependency setup and health checks — Plugin manifests can declare npm, pip, and external runtime dependencies; plugin install and plugin reinstall can provision them with approval, and plugin check reports package, binary, env, and config health for local plugins.

Changed

• Admin console tools and gateway UX — /admin/tools now shows only live built-in and enabled plugin tools, admin tables support click-to-sort headers, the Tools view labels usage as Invocations, and the Gateway page adds a managed restart action with clearer restart-state handling.
• Plugin install ergonomics and memory routing — Local installs accept bare repo plugin ids, prefer plugin-local executables after dependency setup, reuse the standard approval flow for dependency installers, and the bundled MemPalace provider keeps built-in memory active while switching recall to CLI or MCP-backed routes as available.
• Discord concierge approvals — Discord urgency prompts now use native buttons that resume the pending request directly, disable after selection, and keep normal progress reactions visible while the resumed run executes.

Fixed

• Built-in email config reloads — Gateway config changes to built-in email transport settings now restart the email integration automatically so SMTP and IMAP updates apply without a full gateway restart.
• Timezone-aware reminder prompts — Absolute cron reminder guidance now uses offset-bearing one-shot timestamps that mirror the user’s timezone instead of defaulting to Z timestamps in prompt examples.
• Provider and dependency safety — Non-Anthropic provider requests stop sending stray maxTokens, Anthropic requests always get a discovered or fallback limit, and plugin dependency probes no longer execute through a shell while reinstall plans are recalculated from the installed plugin directory.

Contributors

Core

• @furukama — Honcho and MemPalace memory integrations, plugin dependency install/check flows, admin tools and gateway UX improvements, Discord concierge approvals, safer runtime handling, and release prep

All Contributors

@furukama

---

Full Changelog: v0.12.1...v0.12.2

HybridClaw v0.12.1

Release Date: April 9, 2026

The channel-operations and operator-UX release — browser-based channel management in the admin console, remote-access runbooks for SSH and Tailscale, explicit email thread-header control, secret-backed email setup, unified local help, and cleaner secret-handling across TUI and auth flows.

---

Highlights

• Admin console channel operations — Added an /admin Channels workspace with a transport catalog, browser-based editors for Discord, WhatsApp, email, Microsoft Teams, and iMessage, managed secret fields for channel credentials, and live WhatsApp pairing QR display.
• Remote-access runbooks — Added maintainer docs for reaching /chat, /agents, /admin, and remote CLI/TUI clients through SSH tunnels, host-managed Tailscale, and a persistent macOS LaunchAgent tunnel pattern while keeping the gateway bound to loopback.
• Explicit email thread replies — The message tool/API and the repo-shipped brevo-email plugin now accept explicit inReplyTo and references Message-ID headers so outbound replies can attach to existing external mail threads when needed.

Changed

• Secret-backed email transport config — Email setup and runtime config now support email.password as a SecretRef-backed field, and hybridclaw channels email setup keeps stored EMAIL_PASSWORD secrets referenced from config instead of falling back to plaintext values.
• Local operator docs and help surfaces — Refreshed the README, maintainer docs, channel setup guide, FAQ, and public docs landing page for 0.12.1, and unified TUI plus embedded web /help output through the shared command registry so command listings stay alphabetized and surface-aware.

Fixed

• TUI sandbox preflight — hybridclaw tui now follows the sandbox mode reported by a reachable gateway, avoiding unnecessary container rebuild checks when the running gateway is already in host mode and vice versa.
• Secret prompt recovery — Hidden secret prompts now restore terminal state correctly after earlier readline prompts across channel setup and auth flows.
• Credential status masking — auth status surfaces now report sensitive credentials as configured instead of echoing partial tokens, keys, or passwords.

Contributors

Core

• @furukama — admin channel operations, remote-access docs, release docs refresh, explicit email threading headers, secret-backed email setup, unified local help, and TUI/auth secret-handling fixes

All Contributors

@furukama

---

Full Changelog: v0.12.0...v0.12.1

HybridClaw v0.12.0

Release Date: April 8, 2026

The eval-and-knowledge release — managed local benchmark workflows, nightly dream memory consolidation, browser-based skill authoring, per-agent Brevo email routing, and new persistent knowledge skills for long-lived research and note systems.

---

Highlights

• Managed local eval workflows — Added top-level hybridclaw eval plus local /eval, loopback OpenAI-compatible eval environment helpers, managed tau2 lifecycle commands, and a native terminal-bench-2.0 runner with detached logs and progress reporting.
• Dream memory consolidation controls — Added local dream on|off|now commands for nightly workspace-memory consolidation, startup catch-up after downtime, and on-demand cleanup summaries in TUI and web sessions.
• Skill and plugin expansion — Added admin skill creation and ZIP upload flows, the repo-shipped brevo-email plugin for per-agent email addresses, and bundled llm-wiki and zettelkasten skills for persistent research and knowledge maintenance.

Changed

• Skill catalog and operator UX — Normalized skill category metadata across bundled and community skills, grouped skill list output, and refreshed TUI/admin skill management surfaces around availability, review, and authoring.
• Scheduler and console review flow — Improved one-shot scheduler board review so outputs and state are surfaced more reliably, while tightening compact mobile navigation behavior in the embedded console.
• Release-facing docs and landing copy — Refreshed the README, reference docs, skills/plugin docs, and public landing page so the shipped 0.12.0 surfaces match the current runtime and packaging state.

Contributors

Core

• @furukama — eval workflows, dream memory controls, admin skill management, skills/catalog improvements, bundled knowledge skills, release docs, and supporting reliability work
• @snoller — console and mobile UX improvements that shipped in the 0.12.0 release window

All Contributors

@furukama
@snoller

---

Full Changelog: v0.11.0...v0.12.0

HybridClaw v0.11.0

Release Date: April 6, 2026

The gateway-and-operator workflow release — an OpenAI-compatible local gateway API, approval allowlists, a dark-mode admin console, richer web chat controls, clearer channel setup docs, safer packaged runtime pulls, and automated npm publishing for tagged releases.

---

Highlights

• OpenAI-compatible local gateway API — Added loopback-scoped /v1/models and /v1/chat/completions endpoints with streaming responses and usage reporting so local tools can talk to HybridClaw through an OpenAI-compatible surface.
• Approval allowlists and approval UX cleanup — Added workspace approval allowlists, /approve always, and tighter approval wording and replay handling across chat, TUI, and gateway flows.
• Dark-mode admin console and richer web chat controls — Added console dark mode, reusable dropdown UI, cleaner icon extraction, slash-command suggestions in web chat, and improved mobile/browser behavior.

Changed

• Gateway lifecycle and packaged runtime behavior — Improved gateway start/restart and graceful container replacement, reduced duplicate refresh logging, and preferred public runtime image pulls for packaged installs.
• Operator documentation and setup flow — Added channel setup how-tos for Discord, email, WhatsApp, iMessage, and Microsoft Teams, refreshed release-facing docs and README quick start copy, and aligned console dark theme styling with the docs shell.
• Release publishing automation — Added npm publish-on-release automation, renamed the release workflow, and switched npm publishing to OIDC trusted publishing with better pre-release dist-tag handling.

Fixed

• Startup and update guidance — Fixed gateway startup diagnostics, provider auth/model guidance, and post-update restart reminders so local recovery steps are clearer.
• Browser/runtime cleanup reliability — Hardened browser daemon cleanup, kept host browser runtime availability intact, and treated daemon termination as best-effort.
• Config, health, and import edge cases — Fixed runtime config revision synchronization, removed stale imageTag health output, added ClawHub base URL overrides, and retried skill imports more safely after HTTP 429/503 responses.

Contributors

Core

• @furukama — OpenAI-compatible gateway API, gateway lifecycle/runtime improvements, operator docs refresh, release packaging, and README/docs updates
• @maxnoller — dark-mode admin console, dropdown/icon work, web chat/mobile UX, ClawHub import overrides, and release workflow automation
• @snoller — approval allowlist and approval-flow improvements, alias handling, and channel setup documentation

All Contributors

@furukama, @maxnoller, @snoller

---

Full Changelog: v0.10.0...v0.11.0

HybridClaw v0.10.0

Release Date: April 1, 2026

The runtime-secrets and migration release — encrypted credential storage, SecretRef-backed authenticated API calls, OpenClaw and Hermes Agent import flows, first-class llama.cpp support, and smoother local-only onboarding and model-selection guidance.

---

Highlights

• Encrypted runtime secret store — Runtime credentials now live in an encrypted ~/.hybridclaw/credentials.json store with per-secret AES-256-GCM encryption, owner-only permissions, separate master-key sourcing, and automatic migration from legacy plaintext secret files.
• Secret-backed API auth and SecretRefs — Added named /secret storage, SecretRef-backed runtime config fields, and gateway-side http_request auth injection so models can call authenticated APIs without seeing plaintext tokens in prompts or persisted tool payloads.
• OpenClaw and Hermes Agent migration commands — Added hybridclaw migrate openclaw and hybridclaw migrate hermes to preview or import compatible workspace files, config, model settings, and optional secrets into HybridClaw agent workspaces.
• Expanded local-provider setup — Added first-class llama.cpp support, optional model ids for hybridclaw auth login local, and onboarding paths that let local-only operators skip remote-provider auth entirely.

Changed

• Runtime secret loading model — Secret reads now prefer explicit environment overrides and otherwise resolve values from the encrypted store on demand instead of broadly mirroring decrypted values into ambient process.env at startup.
• Secret persistence boundaries and docs — Reserved non-secret runtime config names are excluded from encrypted secret migration and rejected by the /secret surface, while the README, docs shell, FAQ, and runtime docs now reflect the encrypted-secret and SecretRef model.
• Local setup and migration UX — Local backend login flows, startup guidance, and migration reporting are clearer around model selection, local-only onboarding, and repeatable imports into non-main agents.

Fixed

• Startup onboarding loops — Gateway and TUI startup no longer keep re-triggering onboarding once trust acceptance, local-provider setup, or existing credentials already satisfy runtime prerequisites.
• TUI model guidance for local backends — Startup and slash-menu model guidance now give clearer next steps when a local backend is enabled without a selected model.
• Secret migration edge cases — Runtime secret migration now avoids reserved non-secret keys, isolates legacy migration handling, and preserves clearer failure behavior around legacy-home imports and startup secret bootstrap.

Contributors

Core

• @furukama — encrypted runtime secret store, SecretRef-backed API auth, OpenClaw and Hermes Agent migration flows, llama.cpp support, onboarding/model-selection fixes, and release-facing docs updates

All Contributors

@furukama

---

Full Changelog: v0.9.8...v0.10.0

HybridClaw v0.9.8

Release Date: March 31, 2026

HybridClaw v0.9.8 adds concierge routing and tracked runtime config revisions, expands agent install and plugin webhook workflows, bundles Sokosumi, and corrects the shipped HybridAI default model to gpt-4.1-mini.

Highlights

• Added configurable concierge routing with concierge info|on|off, concierge model, and concierge profile <asap|balanced|no_hurry> [model] across gateway, TUI, and slash-command surfaces.
• Added tracked runtime config revisions backed by ~/.hybridclaw/data/config-revisions.db, plus hybridclaw config revisions [list|rollback|delete|clear] for audit and rollback workflows.
• Expanded agent install to support direct .claw URLs, session-side installs, and --skip-import-errors, and added plugin-owned inbound webhook routes plus SDK helpers for dispatching normalized inbound messages.
• Added the bundled sokosumi skill for API-key-authenticated agent hires, coworker task creation, job monitoring, and result retrieval.

Changed

• Updated the shipped HybridAI provider default from gpt-5-nano to gpt-4.1-mini, reordered the built-in HybridAI model list so fresh configs and onboarding pick that model first, and added static capability metadata for gpt-4.1-mini.
• Expanded CI and release verification with broader integration and e2e coverage for gateway docs, config reloads, skill resolution, chat APIs, npm installs, Docker runtime checks, and agent container flows.
• Extracted and tightened gateway plugin service boundaries, improving reload and webhook dispatch behavior while keeping the runtime easier to reason about.
• Refreshed the public docs shell, landing page release content, and release-facing docs so README and the browsable manual match the shipped feature set.

Fixed

• Fixed packaged gateway images so repo docs ship correctly at runtime, and restored static docs deep-link fallback behavior on refresh.
• Fixed packaged install/container reliability around pull-only runtime setup and agent image apt cache locking during builds.
• Fixed agent install partial-failure handling, plugin webhook validation/dispatch edge cases, and tracked config revision robustness around route sanitization, duplicate reads, and watcher timer cleanup.

Contributors

Core

• Benedikt Koehler

All Contributors

• Benedikt Koehler
• Maximilian Noller
• sarthi

Full Changelog: v0.9.7...v0.9.8

HybridClaw v0.9.7

Release Date: March 30, 2026

HybridClaw v0.9.7 adds first-class Mistral support, ATIF-compatible trace export, a searchable /docs help surface, smoother web chat streaming, and aligned release-facing documentation for the new patch.

Highlights

• Added first-class Mistral provider support with hybridclaw auth login|status|logout mistral, mistral/... model selection, and discovered catalog metadata in selectors and status output.
• Added export trace [sessionId|all|--all] so operators can emit ATIF-compatible trace JSONL with tool calls, token usage, git context, and attribution metadata.
• Moved the browsable docs shell to /docs, added raw-markdown /docs/agents.md, and bundled hybridclaw-help so HybridClaw product questions can route through the public docs surface before answering.
• Smoothed built-in web chat streaming with batched rendering, NDJSON fallback handling, replay cleanup, and preserved scroll position during live output.

Changed

• Tightened session previews and export help across /sessions, agent cards, slash menus, and other export-related command surfaces.
• Reworked gateway and agent Docker images into clearer multi-stage builds, added a runtime-lite agent target, and added CI Docker preflight checks plus explicit publish targets.
• Refreshed public docs routing and release-facing docs, and added a clearer HybridClaw Cloud path across the landing page.

Fixed

• Prevented local iMessage self-chat polling from replaying attributed-body control/history rows as fresh inbound messages.
• Expanded trace-export redaction coverage for more token types plus emails, IPs, phone numbers, SSNs, credit cards, and high-entropy secrets, while anonymizing runner-home paths.
• Hardened Mistral discovery and container build/publish flows around canonical model handling, dependency pruning, native addon builds, and runtime target selection.

Contributors

Core

• Benedikt Koehler

All Contributors

• Benedikt Koehler
• Maximilian Noller
• Stephan Noller

Full Changelog: v0.9.6...v0.9.7