--- title: "v1.89.0 - Claude Fable 5, A2A Agent Providers & MCP Per-Server Controls" url: "/release_notes/v1.89.0/v1-89-0" canonical_url: "https://docs.litellm.ai/release_notes/v1.89.0/v1-89-0" type: "release-notes" last_updated: "2026-10-09" related: - "/release_notes/v1.89.1/v1-89-1" - "/release_notes/v1.88.5/v1-88-5" --- # v1.89.0 - Claude Fable 5, A2A Agent Providers & MCP Per-Server Controls > Index of all LiteLLM docs: https://docs.litellm.ai/llms.txt ## Deploy this version **Docker** ```bash docker run \ -e STORE_MODEL_IN_DB=True \ -p 4000:4000 \ docker.litellm.ai/berriai/litellm:v1.89.0 ``` **Pip** ```bash pip install litellm==1.89.0 ``` ## Key Highlights `v1.89.0` builds on [`v1.88.0`](/release_notes/v1.88.0/v1-88-0). - **Claude Fable 5** is supported across Anthropic, Bedrock, Azure AI, and Vertex at 1M-token context with adaptive thinking and computer use. - **Agent-to-agent (A2A)** gains two new agent providers - watsonx Orchestrate and LangFlow (with A2A session bridging) - plus OAuth M2M for Databricks Apps agents. - **MCP gateway** adds per-server environment variables with global and per-user scopes, per-server RPM rate limiting for keys and teams, OAuth passthrough with issuer-scoped JWT auth, and `oauth2_flow` persistence on server registration. - **Observability** lands OpenInference rendering parity for Arize/Phoenix (tool calls, cost, passthrough I/O, sessions, multimodal, cache tokens), MCP semantic conventions on the typed OTel v2 spans, and a Galileo logger that uses the ingest-traces API. - **New search and transcription providers** - APISerpent, You.com, and Soniox - join the gateway, alongside the dashboard's migration to fully typed, OpenAPI-generated API clients. --- ### MCP Credential Store
This release lets you securely store per-server credentials for MCP servers directly on the gateway. Define variables once on a server, scoped either as **Instance** (shared across all users) or **Per-user** (each user supplies their own value), and reference them in static headers or authentication using `${VAR_NAME}` syntax (for example, `${DB_PROTOCOL}://${CORP_USERNAME}:${CORP_PASSWORD}@${DB_HOSTNAME}`), letting each user connect their own identity. [Get Started](../../docs/mcp#server-variables) ## New Providers and Endpoints ### New Providers (3 new providers) | Provider | Supported LiteLLM Endpoints | Description | | ------------------------- | --------------------------- | ------------------------------------- | | APISerpent (`apiserpent`) | Search | Web search and deep-search API | | You.com (`you_com`) | Search | You.com web search API | | Soniox (`soniox`) | Audio Transcription | Async speech-to-text (`stt-async-v4`) | ## New Models / Updated Models #### New Model Support (selected) | Provider | Model | Context Window | Input ($/1M tokens) | Output ($/1M tokens) | Features | | -------------- | --------------------------------------------------------------- | -------------- | ------------------- | -------------------- | ------------------------------------------------------------------------- | | Anthropic | `claude-fable-5` | 1,000,000 | $10.00 | $50.00 | Adaptive thinking, computer use, function calling, prompt caching, vision | | Vertex AI | `vertex_ai/claude-fable-5` | 1,000,000 | $10.00 | $50.00 | Same as Anthropic direct | | Azure AI | `azure_ai/claude-fable-5` | 1,000,000 | $10.00 | $50.00 | Same as Anthropic direct | | Bedrock | `anthropic.claude-fable-5` (+ `global.` / `us.` / `eu.` routes) | 1,000,000 | $10.00 | $50.00 | Same as Anthropic direct | | Bedrock Mantle | `bedrock_mantle/openai.gpt-5.5` | 272,000 | $5.50 | $33.00 | Responses API, reasoning, function calling, prompt caching | | Bedrock Mantle | `bedrock_mantle/openai.gpt-5.4` | 272,000 | $2.75 | $16.50 | Responses API, reasoning, function calling, prompt caching | | Azure AI | `azure_ai/kimi-k2.6` | 262,144 | $0.95 | $4.00 | Reasoning, vision, function calling, tool choice | | MiniMax | `minimax/MiniMax-M3` | 512,000 | $0.60 | $2.40 | Reasoning, prompt caching, function calling | | Inception | `inception/mercury-2` (+ `mercury-edit-2`) | 128,000 | $0.25 | $0.75 | Function calling, prompt caching, response schema | Additional model-map additions: fal.ai Nano Banana and Gemini 2.5 Flash Image generation - [PR #29798](https://github.com/BerriAI/litellm/pull/29798); `mistral/ministral-8b-latest` - [PR #29453](https://github.com/BerriAI/litellm/pull/29453); a batch of new Snowflake Cortex model entries (Claude, GPT, Llama, embeddings); `vertex_ai/google/gemma-4-26b-a4b-it-maas`; APISerpent, You.com, and Soniox catalog entries; and a `jp.` regional route for Claude Opus 4.7. #### Features - **[Anthropic](../../docs/providers/anthropic)** - Route future Claude models to the Anthropic provider via pattern matching - [PR #29239](https://github.com/BerriAI/litellm/pull/29239) - Route Claude Opus 4.8 through adaptive thinking - [PR #29702](https://github.com/BerriAI/litellm/pull/29702) - Emit a thinking block for `reasoning_content`-only streaming chunks in the Anthropic adapter - [PR #29600](https://github.com/BerriAI/litellm/pull/29600) - Inline legacy `$ref` defs in tool schemas (Anthropic and Fireworks) - [PR #28646](https://github.com/BerriAI/litellm/pull/28646) - **[Gemini](../../docs/providers/gemini)** - Support `googleSearch` with server-side tools and `googleMaps` JSON schema - [PR #29582](https://github.com/BerriAI/litellm/pull/29582) - Use GA event names for Pipecat 1.3.x compatibility on Gemini realtime - [PR #29662](https://github.com/BerriAI/litellm/pull/29662) - **[Vertex AI](../../docs/providers/vertex)** - Use a user-supplied `api_base` as-is for the Model Garden OpenAI-compatible path - [PR #29530](https://github.com/BerriAI/litellm/pull/29530) - Handle namespace tools and strip `client_metadata` for Codex compatibility on Vertex/Anthropic - [PR #29489](https://github.com/BerriAI/litellm/pull/29489) - **[Azure AI](../../docs/providers/azure_ai)** - Strip tool-level extra fields on a 400 and retry - [PR #29479](https://github.com/BerriAI/litellm/pull/29479) #### Bug Fixes - **General** - Return a 400 (not 500) on Anthropic context overflow, and seed identity on failed auth - [PR #29848](https://github.com/BerriAI/litellm/pull/29848) - Omit the OpenAI `[DONE]` sentinel on google-genai `streamGenerateContent` - [PR #29426](https://github.com/BerriAI/litellm/pull/29426) ## LLM API Endpoints #### Features - **[Batches](../../docs/batches)** - Skip unnecessary batch input-file reads - [PR #29114](https://github.com/BerriAI/litellm/pull/29114) - Resolve credentials correctly when cancelling a managed batch - [PR #29734](https://github.com/BerriAI/litellm/pull/29734) - **Vector Stores** - Resolve vector-store file-list credentials from team deployments - [PR #29739](https://github.com/BerriAI/litellm/pull/29739) - Support an engines URL for Vertex AI Search - [PR #27885](https://github.com/BerriAI/litellm/pull/27885) - Forward per-request params to Vertex AI Search - [PR #29459](https://github.com/BerriAI/litellm/pull/29459) - **Realtime** - Track realtime audio token cost - [PR #29722](https://github.com/BerriAI/litellm/pull/29722) - Allow null transcripts in stream logging payloads - [PR #29625](https://github.com/BerriAI/litellm/pull/29625) - WebSocket connection improvements - [PR #29563](https://github.com/BerriAI/litellm/pull/29563) #### Agents (A2A) - watsonx Orchestrate agent provider - [PR #29410](https://github.com/BerriAI/litellm/pull/29410) - LangFlow agent provider with A2A session bridging - [PR #28963](https://github.com/BerriAI/litellm/pull/28963) - OAuth M2M for Databricks Apps A2A agents - [PR #29586](https://github.com/BerriAI/litellm/pull/29586) - A2A bug fixes - [PR #29566](https://github.com/BerriAI/litellm/pull/29566) ## Management Endpoints / UI #### Features - **Virtual Keys & Auth** - JWT-to-virtual-key mapping - [PR #28510](https://github.com/BerriAI/litellm/pull/28510) - Let internal users view search tools - [PR #29542](https://github.com/BerriAI/litellm/pull/29542) - Expand the all-team-models sentinel in `can_key_call_model` for batch validation - [PR #29746](https://github.com/BerriAI/litellm/pull/29746) - **Dashboard** - Generate dashboard API types from the proxy OpenAPI spec - [PR #29816](https://github.com/BerriAI/litellm/pull/29816) - Centralize proxy base-URL resolution into a tested resolver - [PR #29793](https://github.com/BerriAI/litellm/pull/29793) - Route networking calls through a shared, location-pinned `apiClient` - [PR #29723](https://github.com/BerriAI/litellm/pull/29723), [PR #29806](https://github.com/BerriAI/litellm/pull/29806), [PR #29815](https://github.com/BerriAI/litellm/pull/29815) - Migrate ESLint to flat config and bump `eslint-config-next` to 16 - [PR #29626](https://github.com/BerriAI/litellm/pull/29626) #### Bug Fixes - Use the resolved DB `user_id` for spend on legacy email match (JWT) - [PR #29217](https://github.com/BerriAI/litellm/pull/29217) - Preserve the 401 status for expired JWTs in OTel traces - [PR #29510](https://github.com/BerriAI/litellm/pull/29510) - Stop team BYOK model-name corruption on model edit - [PR #29731](https://github.com/BerriAI/litellm/pull/29731) - Drop a deleted team BYOK model name from `team.models` - [PR #29820](https://github.com/BerriAI/litellm/pull/29820) - Add `default=None` to `LiteLLM_TeamMembership.litellm_budget_table` - [PR #29684](https://github.com/BerriAI/litellm/pull/29684) - Require a new expiration when regenerating an expired key - [PR #29838](https://github.com/BerriAI/litellm/pull/29838) - Render caller-supplied filter options in caller order (LIT-3151) - [PR #29462](https://github.com/BerriAI/litellm/pull/29462) - Make A2A skill tags enterable and validated - [PR #29512](https://github.com/BerriAI/litellm/pull/29512) - Persist the Tools-tab MCP OAuth token to the DB - [PR #29809](https://github.com/BerriAI/litellm/pull/29809) - Route MCP playground auth by OAuth2 mode instead of `token_url` - [PR #29714](https://github.com/BerriAI/litellm/pull/29714) - Stop MCP playground tool calls from sending twice - [PR #29821](https://github.com/BerriAI/litellm/pull/29821) ## AI Integrations ### Logging - **[Arize / Phoenix](../../docs/proxy/logging)** - OpenInference rendering parity: tool calls, cost, passthrough I/O, session/user, multimodal, and cache tokens - [PR #28800](https://github.com/BerriAI/litellm/pull/28800) - **[Datadog](../../docs/proxy/logging#datadog)** - Split oversized batches on a 413 instead of re-queueing forever - [PR #29444](https://github.com/BerriAI/litellm/pull/29444) - **Galileo** - Use the ingest-traces API and the standard logging payload - [PR #29651](https://github.com/BerriAI/litellm/pull/29651) - **OpenTelemetry** - Allowlist `team_metadata` sub-keys promoted to baggage - [PR #29442](https://github.com/BerriAI/litellm/pull/29442) - Add MCP semantic conventions to OTel v2 - [PR #29468](https://github.com/BerriAI/litellm/pull/29468) - Capture 401 error details in management-endpoint spans - [PR #29535](https://github.com/BerriAI/litellm/pull/29535) - Emit the missing MCP span attributes - [PR #29554](https://github.com/BerriAI/litellm/pull/29554) - Emit a guardrail span on passthrough, including when a guardrail blocks - [PR #29552](https://github.com/BerriAI/litellm/pull/29552), [PR #29470](https://github.com/BerriAI/litellm/pull/29470) ### Guardrails - **[Sensitive Data Routing](../../docs/proxy/guardrails/quick_start)** - Route sensitive data to on-premise models - [PR #29531](https://github.com/BerriAI/litellm/pull/29531) ## Spend Tracking, Budgets and Rate Limiting - Strip NUL bytes from spend-log payloads to prevent PostgreSQL `22P05` errors - [PR #29515](https://github.com/BerriAI/litellm/pull/29515) - Scope the session-token team-key budget exemption to a caller-supplied `team_id` - [PR #29641](https://github.com/BerriAI/litellm/pull/29641) ## MCP Gateway - Per-server environment variables with global and per-user scopes - [PR #28917](https://github.com/BerriAI/litellm/pull/28917) - Per-MCP-server RPM rate limiting for keys and teams - [PR #29482](https://github.com/BerriAI/litellm/pull/29482) - Support MCP OAuth passthrough and issuer-scoped JWT auth - [PR #28356](https://github.com/BerriAI/litellm/pull/28356) - Persist `oauth2_flow` on MCP server registration - [PR #29690](https://github.com/BerriAI/litellm/pull/29690) - Clear `allowed_tools` and tool overrides on MCP server edit - [PR #29411](https://github.com/BerriAI/litellm/pull/29411) - Gate `/public/mcp_hub` strictly on `litellm.public_mcp_servers` - [PR #27764](https://github.com/BerriAI/litellm/pull/27764) ## Performance / Loadbalancing / Reliability improvements - Native `/health/drain` preStop hook for graceful shutdown - [PR #29439](https://github.com/BerriAI/litellm/pull/29439) - Disable proxy buffering on streaming SSE responses - [PR #29557](https://github.com/BerriAI/litellm/pull/29557) - Populate `llm_provider` on internal rate-limit errors - [PR #27707](https://github.com/BerriAI/litellm/pull/27707) - Hot-reload `.env` in dev when running with `--reload` - [PR #29783](https://github.com/BerriAI/litellm/pull/29783) - Enable the Helm backend deployment to mount the gateway `config.yaml` - [PR #29605](https://github.com/BerriAI/litellm/pull/29605) - Convert the AWS and GCP Terraform stacks into reusable modules - [PR #28103](https://github.com/BerriAI/litellm/pull/28103) - Terraform GCP: abandon the SQL user on destroy - [PR #29855](https://github.com/BerriAI/litellm/pull/29855); prompt for `image_registry` in the DeployStack one-click - [PR #29852](https://github.com/BerriAI/litellm/pull/29852) - Dependency bumps - [PR #29860](https://github.com/BerriAI/litellm/pull/29860) ## Documentation Updates - Clarify when to create new test files - [PR #29472](https://github.com/BerriAI/litellm/pull/29472) - Remove fixed dimensions from the README hero image - [PR #29496](https://github.com/BerriAI/litellm/pull/29496) - CLAUDE.md nits - [PR #29504](https://github.com/BerriAI/litellm/pull/29504), [PR #29749](https://github.com/BerriAI/litellm/pull/29749) ### PR roll-up by ownership area ``` PRs by ownership area (visible, non-vehicle set; total: 101) - UI / Dashboard: 22 - General Proxy (testing / CI / build): 22 - Models & Providers: 13 - Performance / Reliability: 10 - Logging: 9 - LLM API Endpoints: 8 - MCP: 6 - Auth & Management: 5 - Agents (A2A): 4 - Docs: 4 - Spend / Budgets / Rate Limits: 2 - Models & Providers (new providers): 3 - Guardrails: 1 ``` ## New Contributors - @someswar177 made their first contribution in https://github.com/BerriAI/litellm/pull/26585 - @trexinc made their first contribution in https://github.com/BerriAI/litellm/pull/26597 - @navnitshukla made their first contribution in https://github.com/BerriAI/litellm/pull/26609 - @tanmay958 made their first contribution in https://github.com/BerriAI/litellm/pull/27580 - @samagana made their first contribution in https://github.com/BerriAI/litellm/pull/27810 - @DrishnaTrivedi made their first contribution in https://github.com/BerriAI/litellm/pull/28330 - @brainsparker made their first contribution in https://github.com/BerriAI/litellm/pull/28370 - @icep87 made their first contribution in https://github.com/BerriAI/litellm/pull/28846 - @adriangomez24 made their first contribution in https://github.com/BerriAI/litellm/pull/29097 - @zzw-math made their first contribution in https://github.com/BerriAI/litellm/pull/29325 - @BeginnerRudy made their first contribution in https://github.com/BerriAI/litellm/pull/29392 - @danisalvaa made their first contribution in https://github.com/BerriAI/litellm/pull/29394 - @kapelame made their first contribution in https://github.com/BerriAI/litellm/pull/29412 - @Zhao73 made their first contribution in https://github.com/BerriAI/litellm/pull/29419 - @suleimanelkhoury made their first contribution in https://github.com/BerriAI/litellm/pull/29420 - @aneeshsangvikar made their first contribution in https://github.com/BerriAI/litellm/pull/29427 - @Ar-maan05 made their first contribution in https://github.com/BerriAI/litellm/pull/29483 - @kingdoooo made their first contribution in https://github.com/BerriAI/litellm/pull/29490 - @dan2k3k4 made their first contribution in https://github.com/BerriAI/litellm/pull/29508 - @yanismiraoui made their first contribution in https://github.com/BerriAI/litellm/pull/29522 - @josx made their first contribution in https://github.com/BerriAI/litellm/pull/29532 - @1qh made their first contribution in https://github.com/BerriAI/litellm/pull/29561 - @tin-berri made their first contribution in https://github.com/BerriAI/litellm/pull/29605 - @mak2508 made their first contribution in https://github.com/BerriAI/litellm/pull/29606 - @VANDRANKI made their first contribution in https://github.com/BerriAI/litellm/pull/29620 - @andrey-dubnik made their first contribution in https://github.com/BerriAI/litellm/pull/29621 - @ErRickow made their first contribution in https://github.com/BerriAI/litellm/pull/29646 - @saswatds made their first contribution in https://github.com/BerriAI/litellm/pull/29650 - @Dinesh-Girbide made their first contribution in https://github.com/BerriAI/litellm/pull/29655 - @BWAAEEEK made their first contribution in https://github.com/BerriAI/litellm/pull/29660 - @hectorc98 made their first contribution in https://github.com/BerriAI/litellm/pull/29672 - @abhay23-AI made their first contribution in https://github.com/BerriAI/litellm/pull/29779 ## Full Changelog https://github.com/BerriAI/litellm/compare/v1.88.0...v1.89.0 --- ## 06/10/2026 - New Models / Updated Models: 16 - LLM API Endpoints: 12 - Management Endpoints / UI: 22 - AI Integrations (Logging / Guardrails): 10 - Spend Tracking, Budgets and Rate Limiting: 2 - MCP Gateway: 6 - Performance / Loadbalancing / Reliability improvements: 10 - General Proxy Improvements (testing / CI / build): 22 - Documentation Updates: 4 ## Related pages - [v1.89.1](https://docs.litellm.ai/release_notes/v1.89.1/v1-89-1.md) - [v1.88.5](https://docs.litellm.ai/release_notes/v1.88.5/v1-88-5.md)