Release notes

Each release has its own page with the new features, fixes and links to the source changes. The same notes ship on GitHub Releases, and the full history is in CHANGELOG.md.

​

2026

​

October

• v1.214.0 — chatcli eval, an offline evaluation harness; generate /help and chatcli ‑‑help from one command registry
• v1.213.1 — render coder and agent output in the active language, with English tool markers; build with Go 1.27.2 and golang.org/x/net v0.60.0; join the memory worker when headless…
• v1.213.0 — Sonnet 5.5, Haiku 5.5 and GPT-6.1 Sol across providers, long-context and cache pricing matched to the vendors’ pages
• v1.212.2 — bump go modules and pin patched alpine packages in operator image

​

September

• v1.212.1 — recover probe and alert stream quickly after a rollout
• v1.212.0 — serve the session slash commands in the browser and run templates as turns
• v1.211.2 — headless chat turns reach the live dashboard and are recorded under the model that served them; a bound or cleared live session never refills from the mcp autosave mirror; New…
• v1.211.1 — derive titles from what the user wrote, never from injected context; standalone web binds a fresh session and bound terminals skip the exit autosave
• v1.211.0 — let the assistant propose switching to coder mode, confirmed by the user; authenticate to JWT-secured Instances and type the RS256, issuer, audience and operator credential…
• v1.210.3 — sessions recorded before the plan rename are brought to the new spelling on load and on every run
• v1.210.2 — the required block is a plan, not the model’s reasoning; both spellings are read
• v1.210.1 — a describe’s subject no longer becomes a tool call, and an empty argv never panics a plugin
• v1.210.0 — Opus 5.5 and the September model launches, self-healing Claude Code fingerprint, refusals name their category
• v1.209.0 — a refused turn falls back to a sibling model, and skill evolution shows on the dashboard
• v1.208.2 — a refused turn is named and resent instead of ending the coder session
• v1.208.1 — process windows move on their own and never cover each other
• v1.208.0 — @dash tool lets the agent manage and read the live dashboard; essentials on the live dashboard page; the dashboard in the deck’s terminal look
• v1.207.2 — prune pulse removes the spool of every process that is gone
• v1.207.1 — move a whole kind at once and never let cards overlap
• v1.207.0 — let nodes be dragged into place on the live dashboard
• v1.206.0 — add the dash command and wire the run and LLM request taps; add the live telemetry bus and its disk spool; add the live telemetry dashboard server and page
• v1.205.1 — deliver instructions typed mid-run instead of dropping them
• v1.205.0 — visible hand-off for user logins, attach to the user’s own Chrome and second-tier verbs
• v1.204.0 — attribute every lost prefix to the request region that changed; cache-shaped extraction request and the memory balance in /cost; bind specific claims to tool evidence in every…
• v1.203.0 — refresh the cache entry through long tool calls on every adapter that can; persist the prompt-cache telemetry with the session and expose the write/read ratio; size the…
• v1.202.0 — keep the coder system prompt byte-stable across runs; scope the cache lifetime per tenant and keep Fable entries warm; honest cache accounting in /cost and a per-request cache log
• v1.201.0 — register the compression knobs in the defaults registry and name them in /config; inventory and on-demand curation of every local store
• v1.200.0 — split the workspace block by lifetime and de-duplicate skills across runs; curate the prefix behind a recovery tool, and cache what is stable
• v1.199.3 — the model listing no longer strips the workspace-trust waiver from turns
• v1.199.2 — waive workspace trust so turns keep working on the current CLI; stop advertising security updates on releases that carry none
• v1.199.1 — the prompt and every palette follow the theme instead of forcing white
• v1.199.0 — o valor GITHUB_MODELS sai do enum de provider do fallback do CRD Instance (spec.fallback.providers[].name) e do schema do chart Helm. Uma Instance que ainda o liste no fallback…
• v1.198.0 — the keychain setting starts doing something, on every platform including Windows; the chart can configure what the security page says it can, and the CORS setting reaches the…
• v1.197.0 — a run under a spending ceiling tells the model how much room is left; a passage carries where it came from into the index; per-turn context as a turn-scoped system message
• v1.196.1 — a reload keeps what the client provided and reaches the servers; environment file discovery for editor-spawned servers and Bedrock profile parity
• v1.196.0 — record every LLM request on every surface in the audit trail; explicit Gemini cache resources, storage pricing and auto TTL; rolling conversation breakpoint, extended TTL and…
• v1.195.0 — list the models the account can invoke via the Devin CLI
• v1.194.0 — refresh provider catalogs for September 2026

​

August

• v1.193.1 — bound shadow-git checkpoints with root guard, deadline and circuit breaker
• v1.193.0 — route high-complexity auto plan-first to @taskgraph; @taskgraph DAG orchestrator with executor-reviewer separation; live browser dashboard for graph runs
• v1.192.0 — doom-loop guard for identical repeated successful tool calls; @browser — Chrome real via DevTools protocol, zero deps novas; outline e map — orientação estrutural sem ler…
• v1.191.0 — one-shot chatcli update subcommand for scripts and automation
• v1.190.1 — drop imaging replace that broke go install and add release-binary fallback
• v1.190.0 — first-class GLM Coding Plan support and custom endpoint routing fix
• v1.189.1 — zero image CVEs, GLM-5.3 catalog, /update pinned to detected release
• v1.189.0 — real usage across all providers, /cost subcommands, budget gates and session persistence
• v1.188.0 — add @embed tool for semantic text operations; refresh image and embedding catalogs for Aug 2026
• v1.187.1 — resolve slash token when piped stdin follows the invocation
• v1.187.0 — refresh provider catalogs and add Bedrock Mantle-to-runtime fallback
• v1.186.1 — stop the loose-code guard from swallowing valid agent_call dispatches
• v1.186.0 — squad worker context navigation, recall and policy parity
• v1.185.0 — auto-route coder-mode slash commands from chat to a coder one-shot run
• v1.184.0 — interop matrix v2 — Codex, Gemini, Qwen, Cursor, Windsurf, opencode and Copilot; universal slash-command templates across every surface
• v1.183.0 — make cross-session memory visible and recallable from turn one
• v1.182.1 — graph expansion line in the auto-recall block
• v1.182.0 — leaner context window across chat and coder modes; persisted knowledge graph participating in recall
• v1.181.3 — recover ACP/MCP turns from dead OAuth tokens, orphaned parks and stdin contention
• v1.181.2 — explain the browser-session failure and give the OAuth callback a real page
• v1.181.1 — bound the permission dialog wait and raise the default to ten minutes; always register a fresh OAuth client when the server supports dynamic registration; bound permission…
• v1.181.0 — terminal parity in client permission dialogs and session policy automode
• v1.180.0 — tool subcommand runs built-in tools directly without an LLM turn; route unattended ask policy through client permission dialogs
• v1.179.0 — add Kimi K3 to the Moonshot catalog with its own pricing tier; refresh Devin CLI model roster; cross-surface session continuity for REPL, ACP, MCP and gateway
• v1.178.0 — live type-ahead preview while the coder spinner runs
• v1.177.0 — cross-process /agents via hub, mid-run side commands and skills in headless runs
• v1.176.0 — agent squad with live run registry, work board, squad mail and autonomous delivery playbook

​

July

• v1.175.0 — default to coder mode guidance and scrub Devin reply sentinels
• v1.174.0 — announce new release at boot and resolve Devin CLI from known install dirs
• v1.173.0 — structured session updates, slash commands and IDE permission dialogs
• v1.172.0 — announce new releases in-session and revamp version UX
• v1.171.0 — list all gateway models when endpoint URL is customized
• v1.170.1 — Bedrock schema override conflicts and provider env reload with /config ‑‑reload alias
• v1.170.0 — add Claude Opus 5 to the Anthropic, Bedrock and OpenRouter catalogs; channel-aware auto-update with /update command and staged background mode
• v1.169.5 — resolve provider model at boot for every provider and refresh Gemini and xAI catalogs
• v1.169.4 — catalog-aware max tokens in all dispatch families and application inference profile resolution
• v1.169.3 — patch grpc-health-probe CVEs in image build
• v1.169.2 — session /max-tokens override reflects in agent, coder and config
• v1.169.1 — Bedrock compaction budget fallbacks and bracket-style tool call parsing
• v1.169.0 — BEDROCK_MODEL selects the default model
• v1.168.1 — control plane uses the corporate HTTP client and bounded listing; invoke Opus 4.8/4.7 and Sonnet 4 through global inference profiles
• v1.168.0 — custom endpoint overrides for corporate DNS and VPC
• v1.167.0 — proactive cross-session recall with tolerant search and titles
• v1.166.1 — preserve delegate prompt and show routed model in turn spinner
• v1.166.0 — capture user messages typed while parked as resume directives
• v1.165.1 — close native tool_call batches on park and pair results positionally
• v1.165.0 — MCP session autosave by default and bounded machine-session lifecycle
• v1.164.0 — derive rollups from the episodic record; episodic timeline with temporal recall; proactive auto-recall in index mode
• v1.163.0 — full ChatCLI experience parity over MCP
• v1.162.1 — re-arm park polling in place to survive rescheduling
• v1.162.0 — @model tool — AI-driven model/provider routing with tiered catalog and delegate
• v1.161.0 — re-export connected MCP servers’ tools on the mcp-server surface
• v1.160.2 — remove writer that broke the completion dropdown; welcome commit hash
• v1.160.1 — repair coder mode input, path validation, prompt rendering and version info
• v1.160.0 — add ui/kit presentation components and fix help header styling; adopt the sóbrio borderless treatment across chat, agent and coder; align command surfaces on kit padding, rules…
• v1.159.0 — OAuth 2.1 authorization for remote MCP servers
• v1.158.0 — DEVIN provider — local Devin CLI as LLM transport; full MCP + ACP server surface — every tool, harness routing, skills as prompts
• v1.157.0 — GPT-5.6 family (Sol/Terra/Luna) + Codex client headers fix
• v1.156.0 — dynamic tool discovery on tools/list_changed notifications
• v1.155.0 — adaptive payload-cap recovery, bounded prompt floor and zero-loss compaction
• v1.154.0 — add @api-explorer builtin for end-to-end API reconnaissance
• v1.153.0 — match multi-word triggers in any word order; re-activate skills mid-loop in agent and coder modes
• v1.152.0 — profile lifecycle, self-healing, chat persistence and deep user expertise layer
• v1.151.1 — parse the flattened flag argv across builtin tools
• v1.151.0 — expose MCP tools through the @tools catalog
• v1.150.0 — defer the tool catalog — index plus on-demand @tools describe; harden CCR reversibility and add runtime profiles; structured HTTP client for the agent as the @http tool
• v1.149.2 — brief every MoA participant like a chat turn and grant read-only tools
• v1.149.1 — route Fable 5 via Mantle and canonicalize profile IDs on the Messages wire
• v1.149.0 — Claude Sonnet 5 + GLM-5.2 + Bedrock Fable 5/Sonnet 5 support

​

June

• v1.148.0 — interactive grap-style, graph in agent, coder and chat
• v1.147.0 — live telemetry in agent/coder + on-demand CCR curation
• v1.146.1 — expand env vars and tilde in file tool paths
• v1.146.0 — self-evolution engine, in-core knowledge graph and memory quality
• v1.145.1 — keep @recall output full through compaction; synchronize JobStatus reads against concurrent transitions
• v1.145.0 — native context compression, image compression and output-token reduction
• v1.144.0 — add @diagram builtin for embedded Graphviz rendering; add selectable system/embedded Graphviz backend to @diagram
• v1.143.1 — refresh model catalog and pin image HTTP calls to HTTP/1.1
• v1.143.0 — multimodal vision input + multi-provider image editing
• v1.142.1 — SIGINT during an operation must not cancel the session
• v1.142.0 — ingest code and infra repos into knowledge bases
• v1.141.1 — re-chunk knowledge corpora on update and harden docs-flatten envelope
• v1.141.0 — full @context parity + @registry-tags sort/last
• v1.140.0 — autonomous knowledge-base management via @context; add @registry-tags and @wikipedia builtins
• v1.139.0 — pure-Go OGG/Opus decoding and actionable ffmpeg guidance
• v1.138.0 — bring @docs-flatten in as a builtin tool
• v1.137.0 — global corporate TLS trust overrides for all outbound clients
• v1.136.2 — stop StackSpot from falling back to the tiny 50K context default; offer the knowledge mode after /context ‑‑mode
• v1.136.1 — bound knowledge retrieval cost by query, not corpus size; clamp the spinner to one terminal row and give @knowledge a short call label
• v1.136.0 — @knowledge tool — pull-side retrieval for agent/coder; knowledge mode — index-card attach + keyless hybrid retrieval
• v1.135.0 — embedded offline Whisper STT via sherpa-onnx
• v1.134.0 — render JS pages via headless browser escalation; keyless fallback providers + Claude Fable 5 in catalog
• v1.133.0 — semantic passage retrieval for /context attach; blend cosine, lexical and temporal signals in retrieval ranking
• v1.132.3 — retry with curl User-Agent on gateway 401/407
• v1.132.2 — authenticate through corporate proxies with special-char credentials
• v1.132.1 — use browser User-Agent to avoid bot blocks
• v1.132.0 — voice replies, embedded TTS and memory resilience
• v1.131.0 — Z.AI + MiniMax image backends, model-aware auto, and /model-image
• v1.130.0 — proactive messaging, productivity skills pack, and multimodal agent tools
• v1.129.0 — voice messages (transcription) + conversational replies
• v1.128.1 — stable-first cache ordering + pull-first memory mode
• v1.128.0 — add interactive ask_user tool for agent, coder and chat

​

May

• v1.127.0 — interactive command palette overlay
• v1.126.0 — gateway runtime model, /model + /max-tokens, 9 new themes
• v1.125.0 — unified color theme system + indentation-safe rendering
• v1.124.0 — add Claude Opus 4.8 with adaptive thinking and fast mode
• v1.123.1 — translate /jobs render output (i18n)
• v1.123.0 — messaging gateway — Telegram, Slack, Discord, WhatsApp, Webhook; cross-channel conversation continuity via shared hub; language server diagnostics via /lsp
• v1.122.5 — hard-break overlong words in result card wrap
• v1.122.4 — wrap streamed tool output to box width
• v1.122.3 — dashboard PostMortem Close button missing and Ack does nothing
• v1.122.2 — close GAP-07 — populate human-action status fields on Issue and PostMortem; swap broken bitnami/kubectl image to registry.k8s.io kubectl
• v1.122.1 — close GAP-06 — Helm CRD upgrade hook and parity tests
• v1.122.0 — close 5 gaps surfaced by 2026-05-23 chaos test
• v1.121.0 — TokenProvider with proactive OAuth refresh and retry-on-401
• v1.120.0 — complete channel push delivery, reactive triggers and persistence
• v1.119.3 — adaptive typewriter pacing so long replies do not crawl
• v1.119.2 — unified bordered response envelope and responsive welcome
• v1.119.1 — restore typewriter effect on assistant responses across chat/agent/coder
• v1.119.0 — cross-mode UI polish — i18n, red errors, lipgloss cards, unified banners
• v1.118.1 — distinct colors for user input and assistant text in compact UI
• v1.118.0 — Claude Code parity tool layer (atomic tools, @todo, multipatch, schema validation, per-tool policies); tool layer enterprise refactor (security UX + parallel orchestration +…
• v1.117.2 — preserve trailing slash on Streamable HTTP endpoint URL
• v1.117.1 — list text/event-stream first in Accept header
• v1.117.0 — add Streamable HTTP transport (MCP 2025-03-26)
• v1.116.0 — add Moonshot (Kimi) provider with thinking-mode support; 15-floor Quality Gate overhaul — fix Floor 3, add 6 floors, redesign table
• v1.115.0 — tier 1-3 server config extensions + autoApprove audit
• v1.114.0 — session-scoped pin for multi-turn auto-injection
• v1.113.1 — restore terminal state after go-prompt so multiline (‑‑-) works
• v1.113.0 — per-server commands + logs ring + manual reload
• v1.112.2 — conserta lifecycle, framing stdio e integração com agent
• v1.112.1 — restore typing and termios after auto-resume
• v1.112.0 — non-blocking park and auto-resume for the ReAct loop

​

April

• v1.111.1 — list and dispatch every model the AWS account has access to
• v1.111.0 — real model windows + env defaults registry for /config
• v1.110.2 — final cache_control boundary + INFO request logs across providers
• v1.110.1 — /schedule quote parser + recurring re-arm with cycle tracking; take per-job lock when snapshotting
• v1.110.0 — loopback OAuth callback for Anthropic w/ branded success redirect
• v1.109.2 — correct openai dim, raise backfill cap, drop dead var
• v1.109.1 — headless agent for /run, /agent, /coder + /jobs clear
• v1.109.0 — durable chronos — /schedule, /wait, /jobs, daemon mode; preflight shell commands against CoderMode policy
• v1.108.0 — force release-please after unparseable commit body
• v1.107.0 — token efficiency pass — caching, early-exit, smart routing, slimmer prompts
• v1.106.1 — /plan dispatch, UX, Claude 4.6 compat, subcommands
• v1.106.0 — autocomplete + i18n for /thinking /refine /verify /plan /reflect; seven-pattern agent pipeline + k8s deps bump; seven-pattern quality pipeline — ReAct, Plan-and-Solve/ReWOO,…
• v1.105.4 — unblock ScaleDeployment=0 with containment opt-in
• v1.105.3 — CRB migration + chart-driven image tag + expanded remediation toolkit
• v1.105.2 — idempotent anomaly creation + broader diagnostic allowlist
• v1.105.1 — stop creating ClusterRoles at runtime (H5 hardening)
• v1.105.0 — hierarchical /config, SearxNG web search, full i18n sweep
• v1.104.0 — compact UX feedback + corporate-proxy payload recovery
• v1.103.1 — stricter credential detection, IMDS control and SSO support
• v1.103.0 — AWS Bedrock provider (Claude + GPT-OSS) with operator/helm support
• v1.102.0 — session sandbox, delegate subagent, webfetch filters
• v1.101.2 — improve model listing for API keys and skip for OAuth
• v1.101.1 — sync operator go.mod/go.sum with root module dependencies
• v1.101.0 — native skills.sh integration with multi-registry support
• v1.100.0 — per-turn model/effort routing for skills and agents
• v1.99.0 — enterprise tool calling, observability, and security hardening; patch libcrypto3/libssl3 CVEs abased 1.98.0
• v1.98.0 — add OpenRouter as LLM provider with full feature parity
• v1.97.4 — upgrade Go stdlib to 1.26.2 to resolve 7 CVEs
• v1.97.3 — patch CVE-2026-34986 in grpc-health-probe
• v1.97.2 — fix Cosign sign step — use tag resolution instead of broken inspect
• v1.97.1 — harden images — distroless, Trivy gate, native multi-arch
• v1.97.0 — hot-reload API keys from both Secret and ConfigMap
• v1.96.1 — sync golang.org/x/time to v0.15.0 matching main module
• v1.96.0 — enterprise security hardening — all 48 audit findings
• v1.95.0 — add built-in overrides with shadow + automatic fallback
• v1.94.1 — skip multi-line block content when injecting changelog annotation
• v1.94.0 — add cosign signing, values schema, and chart documentation
• v1.93.3 — add GHCR login step to publish workflow using ORAS
• v1.93.2 — add version consistency and set strict mode in chart packaging workflow
• v1.93.1 — update workflow to split ArtifactHub metadata per chart and enhance packaging process
• v1.93.0 — add ZAI and MiniMax LLM providers
• v1.92.0 — auto-sync version to docs on release
• v1.91.0 — add ArtifactHub integration and enrich chart metadata

​

March

• v1.90.1 — deduplicate tool calls, wait for user input, add multiline
• v1.90.0 — add GitHub Models marketplace provider
• v1.89.1 — native function calling, compact UI, parser robustness
• v1.89.0 — add MCP client mode, hooks system, web tools, cost tracking, worktrees, channels, and path-specific rules
• v1.88.3 — add mode awareness to system prompts and AI response language support
• v1.88.2 — create parent directory for custom HISTORY_FILE path
• v1.88.1 — resolve shell via COMSPEC/SHELL env vars instead of hardcoded cmd.exe
• v1.88.0 — add REST API authentication with hot-reload and dashboard login
• v1.87.0 — add node monitoring, ReAct loop, approval integration, agentic audit trail
• v1.86.0 — add postmortem developer feedback, fix success rate, harden dashboard security
• v1.85.1 — fix double multiplication in success rate and calculate SLA compliance
• v1.85.0 — wire all orphaned components — zero gaps between code and docs
• v1.84.0 — multi-runbook support — different root causes generate different runbooks
• v1.83.0 — AI validates candidate runbook before execution
• v1.82.0 — add Runbooks tab and configurable auto-refresh control
• v1.81.2 — add aiops field to Instance CRD schema in both Helm charts
• v1.81.1 — remove extra argument in RecordIssueResolved call
• v1.81.0 — configurable resilience, auto-resolve, and dedup invalidation
• v1.80.0 — add Google AI (Gemini) provider support and configuration options; add gRPC server mode, K8s watcher, and comprehensive documentation; add support for OpenAI Responses API and…
• v1.79.0 — AI Insights & Remediations API, dashboard views, and Helm chart version sync
• v1.78.0 — AI Insights & Remediations API, dashboard views, and Helm chart version sync; manual resolve for escalated incidents via API and dashboard
• v1.77.2 — retry Status().Update() on conflict with PostMortemReconciler
• v1.77.1 — PostMortem population, Review/Close state, Audit filters
• v1.77.0 — expand remediation to all K8s workload types (54 actions)
• v1.76.1 — PostMortem standard mode, dashboard overview, audit trail, model lowercase, sortable tables
• v1.76.0 — add provider fallback chain support to Instance CRD
• v1.75.4 — replace wildcard RBAC in superadmin role with explicit resources
• v1.75.3 — rename RBACManager ClusterRoles to avoid Helm name collision
• v1.75.2 — Helm image tag, REST API field mappings, and dashboard display
• v1.75.1 — correct GOOGLE_AI_API_KEY to GOOGLEAI_API_KEY
• v1.75.0 — add dedicated Helm chart for the AIOps operator
• v1.74.0 — automatic rollback engine with structured snapshots
• v1.73.0 — advanced AIOps remediation engine with deep diagnostics
• v1.72.2 — regenerate deepcopy and fix Docker build errors
• v1.72.1 — add missing COPY directives for channels and web packages in Dockerfile
• v1.72.0 — AIOps Platform v2 — enterprise-grade incident management
• v1.71.0 — add dynamic model listing from provider APIs with autocomplete
• v1.70.2 — use detectProjectDir for bootstrap and fix nil response panic
• v1.70.1 — check all profile fields before showing empty message
• v1.70.0 — overhaul memory system with structured storage, smart retrieval and compaction
• v1.69.1 — update default ENV value to prod in documentation and code
• v1.69.0 — refactor/audit remediation complete
• v1.68.5 — add automatic OAuth token refresh and retry on 401 errors
• v1.68.4 — resolve Anthropic OAuth token exchange failures
• v1.68.3 — prevent parallel progress display from corrupting security prompts
• v1.68.2 — add live progress display for multi-agent parallel dispatch
• v1.68.1 — memory worker saving, rewind input, memory autocomplete, NOTHING_NEW filtering
• v1.68.0 — implement documented env vars for bootstrap, memory, and worker config
• v1.67.0 — add memory system, conversation control, and context injection optimization
• v1.66.0 — add provider fallback chain, native tool use, MCP, message bus, and deployment wiring
• v1.65.6 — improve mode context sharing, response spacing and stdin handling
• v1.65.5 — replace SetReadDeadline with cross-platform stdin polling
• v1.65.4 — resolve handshake failure when switching from OpenAI to ClaudeAI
• v1.65.3 — add HTTP/2 support to uTLS Chrome transport
• v1.65.2 — add scoped history management with cross-mode context a…
• v1.65.1 — fix truncated command in security check and raw XML tags in response card

​

February

• v1.65.0 — add interactive local/remote session routing when connected to server
• v1.64.0 — parallel security policies, enhanced prompt UI, smart paste & keyboard navigation
• v1.63.0 — add 6 specialized built-in agents and fix worker LLM client resolution
• v1.62.1 — add project-level agent/skill precedence and activate SetProjectDir
• v1.62.0 — custom persona agents as dispatchable workers + orchestration improvements
• v1.61.0 — add multi-agent orchestration system with parallel workers
• v1.60.0 — remote resource discovery, builtin @coder plugin, paste detection & Windows fixes
• v1.59.6 — stop false-positive remediation loop and enrich AI context
• v1.59.5 — resolve LLM provider from Instance CR instead of hardcoding
• v1.59.4 — add agentic AI remediation and PostMortem CRD
• v1.59.3 — grant Runbook CRD create/update/patch permissions to operator
• v1.59.2 — add remediation intelligence with runbook-first flow, K8s context enrichment and retry re-analysis
• v1.59.1 — resolve RBAC escalation, readonly fs, gRPC keepalive, CRD schema and naming bugs
• v1.59.0 — add keepalive, round-robin LB and headless Service for multi-replica support
• v1.58.0 — AIOps platform with autonomous remediation
• v1.57.2 — update Dockerfiles to Go 1.25 matching go.mod requirement
• v1.57.1 — add GHCR Docker image publishing and update image defaults
• v1.57.0 — multi-target K8s watcher with Prometheus scraping, Operator CRD, and Helm support
• v1.56.1 — harden against shell injection, timing attacks, and infra misconfigs
• v1.56.0 — inject K8s watcher context into agent/coder modes and add interactive /watch start/stop
• v1.55.0 — add gRPC server mode, K8s watcher, and comprehensive documentation
• v1.54.1 — update Anthropic OAuth flow with PKCE and scope adjustments
• v1.54.0 — enforce stricter file and directory permissions across CLI components
• v1.53.4 — improve plugin installation handling for GitHub/GitLab URLs
• v1.53.3 — improve logger error handling and synchronous processing
• v1.53.2 — replace bufio.Scanner with buffered reader for improved stream handling
• v1.53.1 — improve error handling, JSON compacting, and parsing logic in /coder mode
• v1.53.0 — improve error handling, JSON compacting, and parsing logic in /coder mode
• v1.52.0 — add project-specific policy files and enhanced coder mode customizations
• v1.51.1 — clean JSON examples
• v1.51.0 — production-ready @coder and docs
• v1.50.3 — comment out cache control for text response in ClaudeAI client
• v1.50.2 — update OAuth support for ClaudeAI client with streaming response handling
• v1.50.1 — docs(auth): update documentation for OAuth authentication support
• v1.50.0 — implement OAuth2 login flow for OpenAI Codex/GPT Plus with ChatGPT backend support
• v1.49.0 — remove outdated architecture and feature documentation

​

January

• v1.48.2 — add ‑‑raw flag to release prep workflow and bump model
• v1.48.1 — add ‑‑raw flag, enforce tool-call syntax, and update release workflow to gpt-5.2
• v1.48.0 — add Agents customizable feature and frontend-design skill with detailed UX principles and audit script, tokenizer to mode agent and coder
• v1.47.8 — refactor plugin execution with streaming support and enhance error handling
• v1.47.7 — toolcall parser and terminal visuals
• v1.47.6 — improve terminal handling, ANSI reset logic, and add robust unit tests
• v1.47.5 — enhance argument sanitization and add normalization improvements
• v1.47.4 — add support for customizable history file path and improve default paths
• v1.47.3 — improve executable validation and shell logic for cross-platform compatibility
• v1.47.2 — correct escaping and trimming issues in argument parsing
• v1.47.1 — improve line continuation normalization and sanitize tool arguments
• v1.47.0 — add task tracking system and Claude Sonnet 1M tokens beta support
• v1.46.1 — update title and description for governance clarification
• v1.46.0 — add robust security and governance system for /coder mode
• v1.45.0 — improve /coder mode behavior and enable batch tool execution
• v1.44.1 — enforce single-line args in /coder mode tool calls
• v1.44.0 — enhance documentation auto-generation and introduce /coder mode support

​

2025

​

December

• v1.43.8 — improve context handling, parsing logic, and cancellation support
• v1.43.7 — refactor agent mode to streamline logic and improve markdown handling
• v1.43.6 — replace shlex with custom multiline parser for tool args handling
• v1.43.5 — refine coder mode prompts and plugin filtering rules
• v1.43.4 — add HTML decoding and improved multiline command handling
• v1.43.3 — clarify ‑‑encoding base64 requirement in coder mode help text
• v1.43.2 — improve handling of trailing backslashes in tool arguments
• v1.43.1 — add multiline tool args normalization and stricter validation in coder mode
• v1.43.0 — enhance coder mode with role-based context, stricter tool call validation, and execution profile management
• v1.42.0 — add one-shot execution for coder mode and improve translations
• v1.41.1 — enforce reasoning step before tool execution in coder mode and improve error rendering
• v1.41.0 — enhance chatcli-coder with exec command improvements
• v1.40.0 — introduce coder mode and enhance agent UI
• v1.39.3 — enhance Istio configuration and update plugin version

​

November

• v1.39.2 — remove DynamoDB lock implementation and add Pulumi Cloud token support for non-interactive sessions
• v1.39.1 — docs(chatcli-eks): add Pulumi CLI installation instructions to README
• v1.39.0 — add chatcli-eks plugin for EKS cluster deployment
• v1.38.14 — enhance chatcli-docs-flatten with schema generation, better glob matching, and improved CLI
• v1.38.13 — extend chatcli-docs-flatten with Git integration and additional output formats
• v1.38.12 — ensure terminal state is reset after plugin installation confirmation
• v1.38.11 — add chatcli-docs-flatten plugin for Markdown processing and chunk generation
• v1.38.10 — enhance chatcli-kind with improved registry handling and custom Nginx cert-gen options
• v1.38.9 — enhance chatcli-kind with private registry support and Istio improvements
• v1.38.8 — introduce extensive Pulumi-based chatcli-k8s-cloud-refactor example
• v1.38.7 — improve formatting and provide code block examples in agentic-plugins.md
• v1.38.6 — reorganize and extend chatcli-k8s-cloud and chatcli-kind examples
• v1.38.5 — improve context subcommand autocompletions and flag handling
• v1.38.4 — introduce @docker-clean plugin example for managing Docker assets
• v1.38.3 — enhance @kind plugin with new features and improved logging
• v1.38.2 — add new @kind plugin example for Kubernetes cluster management
• v1.38.1 — revise command reference and enhance plugin examples
• v1.38.0 — Implement Agentic AI Engine and Extensible Plugin System
• v1.37.0 — expand ChatCLI documentation with recipes, cookbook index, and advanced workflows
• v1.36.2 — update hero image path and adjust CTA button URL in index.md
• v1.36.1 — use npm ci for consistent installs and add package.json for ghpages dependencies
• v1.36.0 — update create document gihubpages

​

October

• v1.35.0 — add support for updating contexts and enhance context creation
• v1.34.0 — add deep inspection support for contexts and enhance CLI suggestions
• v1.33.0 — implement context management features and enhance dependencies
• v1.32.0 — add support for custom ignore patterns using .chatignore
• v1.31.2 — add demo GIF for chat CLI
• v1.31.1 — add support for ‑‑agent-id flag and update docs/i18n for StackSpot integration
• v1.31.0 — implement internationalization support and update error messages
• v1.30.0 — refactor realm handling and update StackSpot client configuration
• v1.29.1 — remove init-based global config initialization
• v1.29.0 — add comprehensive command validation and execution framework
• v1.28.5 — add support for Claude Sonnet 4.5 with extended context window and aliases

​

September

• v1.28.4 — streamline CI, PR, and release automation
• v1.28.3 — prevent screen clearing issue in certain terminal providers by refining redraw logic
• v1.28.2 — temporarily disable environment variable check and update workflow secrets
• v1.28.1 — add “thinking” filter with customizable option
• v1.28.0 — implement enhanced UI and commands for Agent Mode
• v1.27.2 — enhance security and customization for Agent Mode
• v1.27.1 — introduce retry logic and customizable backoff for API clients
• v1.27.0 — implement override for max_tokens across clients and CLI
• v1.26.0 — add Ollama provider with support for local models
• v1.25.2 — add command flag suggestions and improve ENV log configuration
• v1.25.1 — correct version output and align CLI/main integration
• v1.25.0 — dynamic max_tokens handling and enhanced command guide
• v1.24.0 — integrate xAI provider with support for Grok models
• v1.23.0 — adiciona gerenciamento de sessões e geração de PR assistida por IA
• v1.22.5 — Ensure command history is saved on graceful exit
• v1.22.4 — simplify dynamic header and footer formatting in AgentMode
• v1.22.3 — add build constraints for platform-specific signal files
• v1.22.2 — split forceRefreshPrompt by platform to improve compatibility
• v1.22.1 — remove unused modules and migrate syscall to x/sys/unix in cli/cli.go
• v1.22.0 — Enhance AgentMode UX and refactor input handling
• v1.21.0 — rollback to v1.19.1
• v1.19.1 — Improve command parsing and resolve failing utility tests
• v1.19.0 — add agent one-shot mode with auto-execution support
• v1.18.1 — (test_stackspot_client): add and enhance unit tests, refactor for configurability
• v1.18.0 — improve test robustness by adding error check for form parsing

​

August

• v1.17.4 — add license headers to all source files
• v1.17.3 — comment out pull_request triggers in CI workflow
• v1.17.2 — comment out pull_request triggers in CI workflow; update version comment for testing flow; restructure workflows for improved release management
• v1.17.1 — add progress tracking and file count during directo…; add progress tracking and file count during directory scanning
• v1.17.0 — enhance cancel operation, model switching, and user contex…; enhance cancel operation, model switching, and user context handling; Enhance CLI Control, Agent Mode, and…
• v1.16.3 — centralize one-shot handling in HandleOneShotOrFatal; correct stdin example syntax in help message; streamline input validation and error messaging
• v1.16.2 — normalize -p/‑‑prompt flags without values in one-s…; normalize -p/‑‑prompt flags without values in one-shot mode
• v1.16.1 — extend one-shot mode to support stdin input
• v1.16.0 — add one-shot mode with prompt flags, provider, and model o…; add one-shot mode with prompt flags, provider, and model overrides
• v1.15.0 — add sensitive data sanitization and enhance configuration …; add sensitive data sanitization and enhance configuration commands
• v1.14.1 — correct update command to include ‘v’ prefix in version string; correct update command to include ‘v’ prefix in version…
• v1.14.0 — refactor message building to avoid prompt duplication and add s…; refactor message building to avoid prompt duplication and add system-level instructions
• v1.13.0 — add Google AI (Gemini) provider support and configuration options
• v1.12.1 — improve version detection and update prompts
• v1.12.0 — add support for OpenAI Responses API and enhance configuration options; introduce centralized catalog for model metadata and refactor LLM utilities
• v1.11.0 — enhance model handling and token limits for ClaudeAI and OpenAI

​

June

• v1.10.8 — melhorando o gestor de versão
• v1.10.7 — Rollback mensagem de systema para comandos api OPENAI ASSISTANT
• v1.10.6 — teste fluxo de sync automático na develop
• v1.10.5 — teste fluxo de sync automático na develop
• v1.10.4 — corrigindo erro no lint, nova linha redundante; nova mensagem de prompt para Assistante API OPENAI
• v1.10.3 — Update message version
• v1.10.2 — corrigir referências e fluxo do workflow de release
• v1.10.1 — Melhorar comparação semântica de versões e gestão de recursos
• v1.10.0 — depois de lint resolvido, removendo comentarios de debug; resolvendo lint sinalizado; adiciona comando para verificar e exibir informações de versão