| Version | Release Date |
|---|
| 0.13.0 | 2026-01-05 |
- Users in an organization can now manage servers by license ID (PRO-655)
- Added split combo selector for “org”/“license” (PRO-659)
- Re-added top level Agents and Tools pages (PRO-658)
- Added ability to manage tools in the console (PRO-548)
- Added ability to manage agents in the console (PRO-549)
- Added support for creating/editing agents in console (PRO-619)
- Support for Responses API to use Agents (PRO-589) - Users can prefix model name with “agent:” to use agents instead of models
- Implemented OpenAI Responses API (PRO-578)
- Support for passing header values into MCP calls over Responses API (PRO-665)
- Add Archia hosted GPT-OSS (PRO-607) as a new provider
- Integrate voice component (PRO-536) with realtime speech support
- Server optional root system prompt (PRO-637) - Allows overriding the Archia default system prompt
- Added authentication to the server (archiad) (PRO-533) with authorization header support
- Internal authorization with locally configured token (PRO-535)
- Added clap commands to server to set agents (PRO-567)
- Console Agent Model List (PRO-649) - Now filters based on entered API keys
- Add pagination to console tables (PRO-677) for organizations, licenses, and members
- Add display of exhausted tokens in billing with button to add more (PRO-449)
- Support function granularity for tools on agent definitions (PRO-529) - Restrict agents to specific functions within MCP tools
- Add request override to allow overriding an agent’s model at request time (PRO-622)
- Consider chat options and chat options overrides for agents (PRO-623)
- Support for user supplied timeout of local tools (PRO-528)
- Adjust responses API and agent definitions to honor ReasoningEffort (PRO-652)
- Track execution time for each MCP call independently (PRO-621)
- Define timing values in metrics (PRO-477) with comprehensive tracking including:
- Request issue time (PRO-591)
- First thinking token response time (PRO-592)
- Last non thinking response time (PRO-593)
- Message complete time (PRO-594)
- First/last tool call tokens (PRO-597, PRO-598)
- Token counts for thinking, non-thinking, and tool calls (PRO-595, PRO-596, PRO-599)
- Tokens per second calculations (PRO-600, PRO-601, PRO-602, PRO-603)
- MCP response time computation (PRO-604)
- Support for Rsyslog (PRO-476) with removal of log in favor of tracing
- Define export of Server logs for third party aggregators (PRO-474)
- LangFuse integration testing (PRO-475)
- Prompt caching for local and Cloud LLMs (PRO-531)
- Support for OpenAI (PRO-564)
- Support for Anthropic (PRO-563)
- Support for vLLM (PRO-566)
- Support for Llama Server (PRO-565)
- Test Archia Engine integration with OpenAI spec providers (PRO-510, PRO-511, PRO-512)
- Ensure Thinking Behavior is correct (PRO-627) across Anthropic (PRO-628), OpenAI (PRO-629), Gemini (PRO-630), and Groq (PRO-631)
- Add concurrency testing for the server (PRO-513) across Groq (PRO-632), OpenAI (PRO-633), Anthropic (PRO-634), and Local (PRO-635)
- Fix API keys validation in Console (PRO-664)
- Console Tools/Agents Bugs (PRO-639) - Fixed list of subtools not displaying and hidden sandbox configs
- Console Marketplace Tool Deletion (PRO-640) - Tool marketplace now correctly updates status from “installed” to “uninstalled”
- Subscription Cancellation Display Bug (PRO-646) - Billing components now update correctly after subscription expiration
- Give correct error message for user already existing (PRO-662)
- Disabled link to dashboard (PRO-661)
- Added Archia favicon.ico (PRO-660)
- Spinner holds open for 2 seconds after tool call on Anthropic models (PRO-562)
- Tools registered via Responses API cannot be updated (PRO-668) - Added support for chat specific tools
- Handle invalid json string returned by gemini (PRO-555)
- Ensure engine re-loads tools and agents (PRO-556)
- Non-streaming chats don’t complete from test app (PRO-575)
- Chats with tool calls running out of memory with GPT-OSS on test-server (PRO-576) - Fixed local model memory leak
- Ensure streaming + tool calls work for Genai OpenAI calls (PRO-574)
- Fix orchestrator workflows where sub-agent chats close early (PRO-614)
- Fix version check to be < instead of = for software version (PRO-537)
- Catch archia::gpt-oss headers before they leave the engine* (PRO-653)
- Fix openai tests + handle CaptureUsage (PRO-605)
- Engine Metrics Fix (PRO-612) - Thinking metrics and totals for response time now properly tracked
- tool_properties.toml not updated with tool deletion (PRO-642) - Now cleans up entries on deletion
- Define a default timeout for MCPs that do not define one (PRO-505)
- Invite User To Org UX (PRO-675) - Reduced from 2-step to 1-step process
- Console agents/tools dropdown styling (PRO-676) - Consistent styling across dropdowns
- Console placeholders (PRO-666) - Cleaned up license page and removed unnecessary elements
- Increase model dropdown height for second menu (PRO-443) - Better visibility of available models
- Update default timeout for Apple/Shell/Powershell MCP (PRO-558) - Set to 5.5 minutes
- Remove preload of videos to prevent app slowness (PRO-638)
- Make sure there is an email notification to new org members even if they already exist in the system (PRO-663)
- Update documentation site with last several releases of changes (PRO-470)
- Implement chat with agents using the responses api for the test app (PRO-615)
- Final Demo ideas (PRO-551) - Side by side comparison, voice, image upload