# MLflow changelog

> Tracks experiments, packages models and manages their deployment lifecycle.

- Vendor: LF AI & Data
- Category: AI
- Official site: https://mlflow.org
- Tracked by: What's New (https://whatsnew.fyi/product/mlflow)
- Harvested from: GitHub (mlflow/mlflow)
- Entries below: 9 (newest first)

What's New is an index, not a publisher: every entry below links to the vendor's own release notes, which are the authoritative source. Entries are labelled where they are hand-curated sample data, pre-releases, or drawn from a secondary source such as a developer blog.

Reuse: the summaries, labels and curation here are © What's New. Quote freely with attribution and a link back; wholesale republication of the corpus is not permitted — terms: https://whatsnew.fyi/terms. The vendors' own release notes remain their publishers'.

## Releases

### v3.15.1 — MLflow 3.15.1

- Date: 2026-08-03
- Version: v3.15.1
- Original notes: https://github.com/mlflow/mlflow/releases/tag/v3.15.1
- Permalink: https://whatsnew.fyi/product/mlflow/releases/v3.15.1

- **fixed** — Skip env_pack on ARM client images in Model Registry
- **fixed** — Harden version parsing against missing or non-PEP440 versions on Databricks Serverless in Scoring and Tracking

MLflow 3.15.1 is a patch release that includes bug fixes and documentation updates.

###### Bug fixes:

- [Model Registry] Skip `env_pack` on ARM client images (#24762, @qyc)
- [Scoring / Tracking] Harden version parsing against missing/non-PEP440 versions on Databricks Serverless (#24799, @PattaraS)

###### Documentation updates:

- [Docs] Clarify scorer versioning documentation (#24769, @nihalmenon)

### v3.15.0

- Date: 2026-07-31
- Version: v3.15.0
- Original notes: https://github.com/mlflow/mlflow/releases/tag/v3.15.0
- Permalink: https://whatsnew.fyi/product/mlflow/releases/v3.15.0

- **added** — MCP Registry with semantic-versioned configs, promotable aliases, tags, auto-discovered tools, and ready-made connection instructions for Claude Code and .mcp.json
- **added** — MLflow Assistant now supports multiple LLM providers including Claude Code, Codex, and OpenAI-compatible/Gateway endpoints
- **added** — Sharable named views of the Runs table capturing columns, order, widths, filters, and sort
- **added** — Proxy-less artifact upload/download via presigned URLs for direct cloud storage transfers
- **added** — Multi-modal attachments in LLM judges with get_span_image tool to fetch and display images from trace spans
- **added** — Pydantic AI 2.x autologging support
- **added** — Per-endpoint budget policies in the AI Gateway
- **added** — SAP AI Core Orchestration v2 provider with extra_headers support
- **added** — Action recommendation for low-value issue detection result
- **added** — Search traces functionality in TypeScript SDK
- **added** — Per-tool cards with Monaco JSON editor and Format action in the Playground
- **added** — Redesigned Playground Tools settings with Add-tools entry flow
- **added** — Cursor model catalog with Auto and Composer 2.5 pricing
- **added** — Numeric assessment trace filters support
- **added** — mlflow agent setup command for single-prompt Assistant setup with secure API key storage
- **changed** — Assistant now displays live per-session token usage and estimated cost in the composer
- **changed** — UC model artifact download migrated to native temp-creds
- **changed** — UC model-registry client gated on native /api/2.1 endpoints
- **changed** — Tracing UI stopped eagerly prefetching full neighboring traces
- **fixed** — OpenTel parent context propagation in tracing
- **fixed** — Artifact direct download capability checks in UI
- **fixed** — Traces saved-view refresh on load across multiple windows
- **fixed** — UnauthorizedError constructor message
- **fixed** — Fall back to nested spans for Chat Sessions turn content
- **fixed** — Experiment GC cascades for traces and logged models
- **fixed** — Workspace-enabled tracking store crash-loop when default experiment is renamed
- **fixed** — Kubernetes auth token extraction for kubernetes-client 36+
- **fixed** — Assistant config refresh when leaving settings to prevent stale provider indicator
- **fixed** — Issue detection structured output schema
- **fixed** — mlflow.statsmodels flavor with MLFLOW_ALLOW_PICKLE_DESERIALIZATION guard
- **fixed** — Static prefixes preservation in notebook trace renderer URL
- **fixed** — Blank fields in run detail sidebar
- **fixed** — Reasoning content parts normalization for ChatAgentMessage in parse_message
- **fixed** — Evaluation Runs Actions dropdown remaining clickable when no runs are selected
- **fixed** — Truncated assistant streams surfaced as errors instead of silent done
- **fixed** — Experiment grants honored on artifacts-only servers when workspaces disabled
- **fixed** — Get-history-bulk-interval memory bounded by sampling metric history by row
- **fixed** — Artifact authorization enforcement on presigned download route in basic-auth
- **fixed** — Explicit expectations preservation in mlflow.genai.evaluate
- **fixed** — Portkey routing configuration support in the AI Gateway
- **fixed** — Input modal submission errors shown inline instead of transient toast in GenericInputModal

MLflow 3.15.0 includes several major features and improvements

###### Major New Features

- **🧩 [MCP Registry](https://mlflow.org/docs/latest/genai/mcp-registry/)**: A centralized catalog for registering, versioning, and sharing Model Context Protocol servers — with semantic-versioned configs, promotable aliases, tags, auto-discovered tools, and ready-made connection instructions for Claude Code and `.mcp.json`. Manage it from the UI, REST API, or Python!
- **🤖 MLflow Assistant enhancements**: The in-app AI assistant now supports multiple LLM providers (Claude Code, Codex, and OpenAI-compatible/Gateway endpoints) chosen from a single settings page, displays live per-session token usage and estimated cost in the composer, and is easier to set up — `mlflow agent setup` can enable it in one prompt, with API keys stored securely in the Gateway's LLM Connections.
- **🗂️ Sharable table views**: Save named views of the Runs table — capturing columns, order, widths, filters, and sort — and share them by URL.
- **⚡ Proxy-less artifact upload/download via presigned URLs**: Large artifact transfers can now bypass the tracking server and talk directly to cloud storage (e.g. S3) through presigned URLs, cutting server load and timeouts on big files. We fall back to proxied transfer automatically for backward compatibility.
- **🖼️ Multi-modal attachments in LLM judges**: `{{ trace }}` judges created with `make_judge()` can now _see_ images captured in trace spans via a new `get_span_image` tool, which fetches attachments as base64 data URLs and passes them to multimodal models (Anthropic, Gemini, and OpenAI-compatible via litellm) — enabling evaluation of vision tasks, screenshots, and other visual content.

###### Other Assorted Features & Improvements:

- [Tracing / Tracking] Add Pydantic AI 2.x autologging support (#24721, @B-Step62)
- [Tracing / UI] [3/3] Add action recommendation for low-value issue detection result (#24580, @B-Step62)
- [Gateway] Support per-endpoint budget policies in the AI Gateway (#24370, @joelrobin18)
- [Evaluation / Gateway] Add `sap-ai-core:/` provider for SAP AI Core Orchestration v2 with `extra_headers` support (#24356, @shivamxtech)
- [Model Registry] [Model Registry] Migrate UC model artifact download to native temp-creds (#24576, @AknA13)
- [Gateway / Tracing / UI] [2/3] Simplify issue detection start modal (#24579, @B-Step62)
- [Tracing / UI] [1/3] Run issue detection in the background (#24578, @B-Step62)
- [Tracing] [BUG] Fix otel parent context propagation (#24444, @iyashk)
- [Model Registry] [Model Registry] Gate UC model-registry client on native `/api/2.1` endpoints (#24517, @AknA13)
- [Docs / Tracking] Enable artifacts-only mode with workspaces (#24452, @kramaranya)
- [Tracing / UI] Stop eagerly prefetching full neighboring traces in tracing UI (#24411, @DaoDaoNoCode)
- [Docs / Gateway] Authenticate OpenAI-protocol coding agents through the RBAC gateway (#24294, @joshuawong-db)
- [Tracing] feat: add search traces to typescript sdk (#23660, @WUMIKE233)
- [Prompts / UI] Add per-tool cards with a Monaco JSON editor and Format action to the Playground (#24129, @tanghaoji)
- [Prompts / UI] Redesign Playground `Tools` settings with an Add-tools entry flow (#24102, @tanghaoji)
- [] feat: add Cursor model catalog with Auto and Composer 2.5 pricing (#24078, @TomeHirata)
- [Tracing / Tracking] Support numeric assessment trace filters (#23948, @tanghaoji)

Bug fixes:

- [Tracing] Compute `mlflow.llm.cost` for Claude Code plugin traces on Databricks (#24416, @harupy)
- [UI] Fix artifact direct download capability checks in UI (#24748, @B-Step62)
- [Tracing / UI] Refresh traces saved-view on load across multiple windows (#24746, @joshuawong-db)
- [UI] Fix UnauthorizedError constructor message (#24729, @mihir2921)
- [UI] Fall back to nested spans for Chat Sessions turn content (#24703, @MannXo)
- [Tracing / Tracking] Fix experiment GC cascades for traces and logged models

_[Truncated at 4000 characters — full notes: https://github.com/mlflow/mlflow/releases/tag/v3.15.0]_

### v3.14.0

- Date: 2026-06-17
- Version: v3.14.0
- Original notes: https://github.com/mlflow/mlflow/releases/tag/v3.14.0
- Permalink: https://whatsnew.fyi/product/mlflow/releases/v3.14.0

- **added** — One-command agent onboarding with `mlflow agent setup` to install MLflow, set up tracing, and configure coding agents with MLflow skills
- **added** — Durable, low-latency tracing for Claude Code with write-ahead-log to prevent slowing the agent, overwhelming the tracking server, or losing traces
- **added** — Review Queues for traces to assign traces to reviewers and collect structured feedback and ground-truth annotations in the UI
- **changed** — Revamped evaluation dataset UI to browse, inspect, edit, and bulk-manage evaluation dataset records directly in the UI with click-through to source trace
- **added** — Pytest integration for regression testing with `@mlflow.test` marker to write GenAI regression tests and review test history in the UI
- **added** — LLM Playground to iterate on prompts in the browser against AI Gateway endpoints and Prompt Registry versions with settings, tools, and structured output
- **changed** — Change `mlflow.sklearn` `serialization_format` default from `cloudpickle` to `skops`
- **changed** — Change `serialization_format` default to `pt2` for `mlflow.pytorch.log_model` and `mlflow.pytorch.save_model`
- **changed** — Change `serialization_format` default to `skops` in `mlflow.lightgbm` `log_model`/`save_model`
- **added** — Add EvaluationResult.passed and .reason properties for `@mlflow.test` assertions
- **added** — Add shareable review queue URLs with a `startReview` deep link
- **added** — Allow editing a completed review in place in focus mode
- **added** — Add `x-mlflow-run-id` support to OTLP trace ingestion
- **added** — Add `mlflow skills view/list` CLI command
- **added** — Add `MLFLOW_WORKSPACE` support to OSS auth provider
- **added** — Add cached token pricing to Databricks model catalog in Gateway
- **added** — Add `MLFLOW_GENAI_JUDGE_DEFAULT_MODEL` environment variable for evaluation
- **added** — Add rule-based built-in scorers: `RegexMatch`, `PIIDetection`, `ResponseLength`
- **added** — Add Google ADK LLM judge scorers (`Hallucination`, `Safety`, `ResponseEvaluation`)
- **fixed** — Fix `ChrfScore` RAGAS scorer instantiation due to class name mismatch

MLflow 3.14.0 includes several major features and improvements

###### Major New Features

- **🚀 [One-command agent onboarding with `mlflow agent setup`](https://mlflow.org/docs/latest/genai/tracing/quickstart/)**: Install MLflow, set up tracing, and hand your favorite coding agent (Claude Code, OpenAI Codex, or OpenCode) the MLflow skills to instrument your app, all from a single command.
- **⚡ [Durable, low-latency tracing for Claude Code](https://mlflow.org/docs/latest/genai/tracing/integrations/listing/claude_code)**: Roll out Claude Code tracing across a team with confidence: a write-ahead-log keeps it from slowing the agent, overwhelming the tracking server, or losing traces on a network blip or crash.
- **📝 [Review Queues for traces](https://mlflow.org/docs/latest/genai/assessments/review-queues)**: Assign traces to reviewers (or agents) and collect structured feedback and ground-truth annotations in the UI, written straight back onto the trace so they are immediately usable for evaluation.
- **🗂️ [Revamped evaluation dataset UI](https://mlflow.org/docs/latest/genai/datasets/)**: Browse, inspect, edit, and bulk-manage evaluation dataset records directly in the UI, with click-through to the source trace.
- **🧪 [Pytest integration for regression testing](https://mlflow.org/docs/latest/genai/eval-monitor/regression-testing)**: Write GenAI regression tests as plain pytest functions with the `@mlflow.test` marker, gate them in CI, and review test history and per-assertion judge results in the UI.
- **🎛️ [LLM Playground](https://mlflow.org/docs/latest/genai/prompt-registry/playground)**: Iterate on prompts in the browser against your AI Gateway endpoints and Prompt Registry versions, with settings, tools, structured output, and template variables.

###### Breaking Changes

- [Models] Change `mlflow.sklearn` `serialization_format` default from `cloudpickle` to `skops` (#23987, @copilot-swe-agent)
- [Models] Change `serialization_format` default to `"pt2"` for `mlflow.pytorch.log_model` and `mlflow.pytorch.save_model` (#23988, @copilot-swe-agent)
- [Models] Change `serialization_format` default to `"skops"` in `mlflow.lightgbm` `log_model`/`save_model` (#23986, @copilot-swe-agent)

###### Other Assorted Features & Improvements:

- [Evaluation / UI] [3/3] Show regression-test results in the existing eval-run UI (#23985, @B-Step62)
- [Prompts / UI] Add "Save prompt to registry" action to the Prompt Playground (#24021, @B-Step62)
- [Prompts] Prompt Playground (#23273, @TomeHirata)
- [Evaluation] [2/3] Add `EvaluationResult.passed`/`.reason` for `@mlflow.test` assertions (#23869, @B-Step62)
- [UI] Review queues: list the affected queues in the delete-question confirmation (#24002, @kriscon-db)
- [UI] Add shareable review queue URLs with a `startReview` deep link (#23941, @harupy)
- [UI] Allow editing a completed review in place in focus mode (#23967, @kriscon-db)
- [Tracing] Add `x-mlflow-run-id` support to OTLP trace ingestion (#23664, @sanatb187)
- [Evaluation / Tracing] [1/3] Add @mlflow.test pytest marker and assertion framework (#23864, @B-Step62)
- [UI] Improve review queue empty states with onboarding content (#23903, @B-Step62)
- [UI] Add `mlflow skills view/list` CLI (#23907, @joshuawong-db)
- [UI] Improve review queue list: flat layout, sortable columns, status filter (#23902, @B-Step62)
- [Tracing] Add `MLFLOW_WORKSPACE` support to OSS auth provider (#23927, @Nehanth)
- [Gateway] Add cached token pricing to Databricks model catalog (#23901, @TomeHirata)
- [Evaluation] Add `MLFLOW_GENAI_JUDGE_DEFAULT_MODEL` environment variable (#23860, @B-Step62)
- [Evaluation] Wire "Run judge(s)" submission in "Run Eval" in Evaluations Run page to POST /mlflow/genai/evaluate/invoke (#23781, @aaronteo-db)
- [Evaluation] Add rule-based built-in scorers: `RegexMatch`, `PIIDetection`, `ResponseLength` (#22571, @debu-sinha)
- [Tracing] Support Databricks backend in `mlflow agent setup` (#23783, @harup

_[Truncated at 4000 characters — full notes: https://github.com/mlflow/mlflow/releases/tag/v3.14.0]_

### v3.13.0

- Date: 2026-06-01
- Version: v3.13.0
- Original notes: https://github.com/mlflow/mlflow/releases/tag/v3.13.0
- Permalink: https://whatsnew.fyi/product/mlflow/releases/v3.13.0

- **added** — Role-Based Access Control system with reusable roles and workspace-scoped grants
- **added** — Admin UI for managing users, roles, and permissions on self-hosted MLflow
- **added** — Trace retention and auto-archival to move aged trace span data to object storage while keeping traces readable in UI and APIs
- **added** — One-click observability and governance for coding agents including Claude Code, OpenAI Codex, and Gemini CLI through the AI Gateway
- **added** — New engines for MLflow Assistant including local Ollama models, OpenAI Codex CLI, and MLflow AI Gateway endpoints
- **added** — Official production-ready Helm chart for deploying MLflow tracking server to Kubernetes clusters
- **added** — Hermes Agent support to route through AI Gateway and capture end-to-end traces over OpenTelemetry
- **added** — Span log levels with Python-logging-style severity levels and Minimum log level filter in trace UI
- **added** — AI Gateway as a backend for MLflow Assistant
- **added** — Unified per-user permission APIs under /mlflow/users/permissions/*
- **added** — mlflow.genai.test_agent for automated agent stress-testing
- **added** — Raw proxy endpoint at /gateway/proxy/{endpoint_name}/{path}
- **added** — Ollama as assistant provider
- **added** — Prompt promoted to a first-class RBAC resource_type
- **added** — Link entity and LiveSpan.add_link() for OpenTelemetry Span Links
- **changed** — Permission system overhauled into unified Role-Based Access Control model replacing legacy per-resource permission tables, REST endpoints, and client methods
- **changed** — default_permission now acts as a floor rather than an override
- **changed** — Workspace USE grant is now sufficient to create experiments and registered models
- **changed** — Default optimizer used by judge.align() is now MemAlign
- **removed** — MLServer as a pyfunc serving backend and the previously deprecated enable_mlserver option
- **removed** — mlflow autolog claude no longer installs the old Python autolog hook in favor of the official Claude plugin
- **fixed** — Pointing tracking or model registry store at local file-system path now raises an error by default unless MLFLOW_ALLOW_FILE_STORE=true is set

MLflow 3.13.0 includes several major features and improvements

###### Major New Features

- **🔐 [Role-Based Access Control & Admin UI](https://mlflow.org/docs/latest/self-hosting/security/role-based-access-control)**: A full RBAC system with reusable roles and workspace-scoped grants, plus a new web Admin UI for managing users, roles, and permissions on self-hosted MLflow.
- **🗄️ [Trace Retention & Auto Archival](https://mlflow.org/docs/latest/genai/tracing/observe-with-traces/archive-traces)**: Automatically move aged trace span data out of your SQL backend into object storage (e.g. S3) while keeping every trace fully readable in the UI and APIs.
- **🤖 [One-click observability & governance for coding agents](https://mlflow.org/docs/latest/genai/governance/ai-gateway/coding-agents/)**: Onboard Claude Code, OpenAI Codex, or Gemini CLI to the AI Gateway in one click for tracing, usage tracking, budgets, and guardrails.
- **✨ [New engines for MLflow Assistant](https://mlflow.org/docs/latest/genai/getting-started/try-assistant/)**: Run MLflow Assistant on a local Ollama model, the OpenAI Codex CLI, or any MLflow AI Gateway endpoint, in addition to Claude Code.
- **☸️ [Helm chart for Kubernetes](https://mlflow.org/docs/latest/self-hosting/kubernetes-helm)**: An official, production-ready Helm chart for deploying the MLflow tracking server to any Kubernetes cluster.
- **🌐 [Hermes Agent support](https://mlflow.org/docs/latest/genai/governance/ai-gateway/coding-agents/hermes-agent)**: Route the Hermes Agent runtime through the AI Gateway and capture its end-to-end traces in MLflow over OpenTelemetry.
- **🪵 [Span log levels](https://mlflow.org/docs/latest/genai/tracing/app-instrumentation/logging)**: Python-`logging`-style severity levels on spans, with a "Minimum log level" filter in the trace UI to hide low-level noise.

###### Breaking Changes

- The permission system has been overhauled into a unified Role-Based Access Control model. The legacy per-resource permission tables, REST endpoints, and client methods are removed and replaced by roles backed by `role_permissions`, `default_permission` now acts as a floor rather than an override, and a workspace `USE` grant is sufficient to create experiments and registered models. Code that relied on the old per-resource permission APIs must migrate to the new role-based APIs. (#22855, #22859, #22941, #23337, #23379, @PattaraS)
- MLServer is no longer available as a pyfunc serving backend. The previously deprecated `enable_mlserver` option has been removed, so `mlflow models serve` always uses the built-in scoring server. (#23356, @harupy)
- `mlflow autolog claude` no longer installs the old Python autolog hook; Claude Code tracing is now provided by the official Claude plugin, which must be installed separately. (#23339, @B-Step62)
- The default optimizer used by `judge.align()` is now MemAlign, so existing alignment workflows may produce different judges than before unless an optimizer is passed explicitly. (#23254, @veronicalyu320)
- Pointing the tracking or model registry store at a local file-system path now raises an error by default; set `MLFLOW_ALLOW_FILE_STORE=true` to keep using a file-based store. (#22773, @harupy)

###### Other Assorted Features & Improvements:

- [Gateway] Support AI Gateway as a backend of MLflow Assistant (#23559, @B-Step62)
- [UI] Make admin pickers target the workspace they're granting into (#23543, @PattaraS)
- [UI] Bring direct-grant picker to parity with role picker (#23420, @PattaraS)
- [UI] Cherry-pick: Add OpenAI Codex CLI as assistant provider (#22566) (#23517, @B-Step62)
- [Tracing] Support settings.local.json for Claude Code tracing config (#23285, @Gkrumbach07)
- [UI] Add coding-agent endpoint creation flow in AI Gateway UI (#23430, @TomeHirata)
- [Tracking / UI] Unified per-user permission APIs: `grant` / `revoke` / `get` / `list` under `/mlflow/users/permissions/*` (#23247, @PattaraS)
- [Evaluation / Tracing]

_[Truncated at 4000 characters — full notes: https://github.com/mlflow/mlflow/releases/tag/v3.13.0]_

### v3.13.0rc0

- Date: 2026-05-22
- Version: v3.13.0rc0
- Original notes: https://github.com/mlflow/mlflow/releases/tag/v3.13.0rc0
- Permalink: https://whatsnew.fyi/product/mlflow/releases/v3.13.0rc0
- Labels: Pre-release

- **added** — Major overhaul of Role-Based Access Control with unified permission APIs, workspace USE permission for creating experiments and registered models, prompt promoted to first-class RBAC resource_type, and new Admin UI with account widget and Platform Admin pages
- **added** — Claude Code, OpenClaw, Ollama, and OpenAI Codex integrated as first-class assistant providers in the AI Gateway with Claude Code TypeScript plugin and setup wizard
- **added** — End-to-end trace archival across the tracking stack with archival configuration models, OTLP and artifact helpers, and archive-aware retrieval fallback
- **added** — Helm charts for deploying MLflow to Kubernetes clusters with production-ready configuration, ingress, and persistence
- **added** — mlflow.genai.test_agent API for automated agent stress-testing with adversarial input generation and trace review
- **added** — OpenTelemetry Span Links support via LiveSpan.add_link() to connect causally related spans across traces
- **added** — Database replica routing in SQL tracking store for reader/writer instance routing to enable horizontal scaling
- **changed** — Legacy per-resource permission tables collapsed into unified role_permissions table

We're excited to announce MLflow 3.13.0rc0, which deepens agent observability, tightens permissions, and broadens deployment options:

**Major New Features**:

- **RBAC + Admin UI**: Major overhaul of MLflow's Role-Based Access Control — legacy per-resource permission tables collapsed into `role_permissions`, unified per-user permission APIs under `/mlflow/users/permissions/*`, workspace `USE` permission lets users create experiments and registered models, default roles are seeded on workspace creation, `prompt` is promoted to a first-class RBAC `resource_type`, and a new 4-page Admin UI (account widget, `/account` page, Platform Admin pages, backend auth endpoints) opens to workspace managers scoped per their workspace. (#22855, #22857, #22859, #22928, #22929, #22941, #22973, #23086, #23247, #23248, #23337, #23379, @PattaraS)

- **Coding-Agent Tracing as Plugins**: Claude Code, OpenClaw, Ollama, and OpenAI Codex are now wired into the AI Gateway as first-class assistant providers, plus a Claude Code TypeScript plugin with a setup wizard and `settings.local.json` support. The legacy Python autolog hook for `mlflow autolog claude` is replaced by the new official plugin, and a coding-agent endpoint creation flow is now available directly in the AI Gateway UI. (#20414, #22098, #22566, #22717, #23218, #23285, #23339, #23430, #23517, @B-Step62, @joelrobin18, @Gkrumbach07, @SuperSonnix71, @TomeHirata)

- **Trace Archival**: End-to-end trace archival across the tracking stack. Includes archival configuration models, OTLP and artifact helpers, SQLAlchemy archival passes, archive-aware retrieval fallback, plus workspace/experiment/server-level archival settings in the UI. Read archived traces back seamlessly. (#23359, @mprahl)

- **Helm Charts for Kubernetes Deployment**: First-class Helm chart for deploying MLflow to Kubernetes clusters — production-ready configuration, ingress, persistence, and `appVersion` wired to the released MLflow image. Get from `helm install` to a running tracking server without writing your own manifests. (#21973, @WeichenXu123)

- **`mlflow.genai.test_agent` for Automated Agent Stress-Testing**: New API for stress-testing GenAI agents — generate adversarial inputs, replay them through your agent, and review the resulting traces in MLflow. Wires into the existing evaluation flow and assessment APIs. (#22990, @serena-ruan)

- **OpenTelemetry Span Links**: Tracing now supports the OpenTelemetry `Link` entity via `LiveSpan.add_link()`, letting you connect causally related spans across traces. (#22797, @khaledsulayman)

- **Database Replica Routing**: The SQL tracking store now supports reader/writer instance routing for database replicas, so read-heavy MLflow deployments can scale horizontally without overloading the primary. (#22910, @ravidarbha)

Stay tuned for the full release, which will include even more features and bug fixes.

To try out this release candidate, please run:

`pip install mlflow==3.13.0rc0`

### ts/v0.2.0 — TypeScript SDK 0.2.0

- Date: 2026-05-15
- Version: ts/v0.2.0
- Original notes: https://github.com/mlflow/mlflow/releases/tag/ts/v0.2.0
- Permalink: https://whatsnew.fyi/product/mlflow/releases/ts-v0.2.0

Bump several RC TypeScript packages stable version.

### v3.12.0

- Date: 2026-05-05
- Version: v3.12.0
- Original notes: https://github.com/mlflow/mlflow/releases/tag/v3.12.0
- Permalink: https://whatsnew.fyi/product/mlflow/releases/v3.12.0

- **added** — Store multimodal content in tracing spans as artifact attachments with mlflow-attachment:// style URI and rich rendering for PDFs, audio, and images
- **added** — Support for Codex, Gemini, and Qwen coding agent tracing
- **added** — Set guardrails on gateway endpoints to prevent unsafe or non-compliant model inputs and outputs
- **added** — Pagination for traces tab to improve initial load time and page responsiveness
- **added** — Record caller in gateway traces from request headers
- **added** — Portkey as a supported AI Gateway provider
- **added** — Add last_updated_at field to model catalog entries
- **added** — Enable third-party scorer registration in OSS MLflow
- **added** — Auto-start SQL warehouse before V4/V5 MLflow tracing calls
- **added** — Support model_kwargs in DeepEval scorers for LLM parameter control
- **added** — Application Default Credentials auth mode for Vertex AI gateway
- **added** — Share button to trace detail view
- **added** — mlflow.diffusers flavor for diffusion model LoRA adapters
- **added** — Add size limit for trace attachments
- **added** — JSON OTLP encoding support for trace ingestion
- **added** — Presigned upload URL endpoint for S3 artifact uploads
- **added** — Add @mlflow/vercel for better Vercel AI SDK tracing in Databricks UC
- **added** — Guardrail base class and JudgeGuardrail implementation
- **added** — Configurable provider allow/deny list to AI Gateway
- **added** — mlflow db move-resources command to move resources between workspaces

MLflow 3.12.0 includes several major features and improvements

###### Major New Features

- **🖼️ Multimodal Tracing**: Users can now store multimodal content in tracing spans as artifact attachments instead of inline binary data. We've also patched the UI to support the new mlflow-attachment:// style URI, with rich rendering available for PDFs, audio, and images.
- **🤖 Codex, Gemini, Qwen coding agent tracing support**: Similar to our Claude Code tracing integration, we've now added support for the Codex, Gemini, and Qwen coding agent platforms as well!
- **🛡️ Gateway guardrails**: You can now set guardrails on your gateway endpoints to prevent unsafe or non-compliant model inputs and outputs. Try it out in the MLflow UI!
- **⚡ Trace table pagination**: The traces tab is now paginated, rather than fetching all traces up to a limit of 1000. This improves initial load time, and makes the page feel more responsive overall.

###### Breaking Changes

- [Scoring] Deprecate `enable_mlserver` in pyfunc serving backend (#22994, @B-Step62)

###### Other Assorted Features & Improvements:

- [UI] Add coding agents section to AI Gateway quick start (#23006, @TomeHirata)
- [Tracing] feat: record caller in gateway traces from request headers (#22926, @TomeHirata)
- [] Run guardrails on passthrough endpoints; skip `response_format` for non-chat payloads (#22856, @TomeHirata)
- [] Gateway: preserve client auth header for subscription-based CLI tools (`claude-cli`, `Codex-Desktop`, `GeminiCLI`) (#22915, @TomeHirata)
- [Tracing] Expose Codex and Qwen Code hooks as installable CLI binaries (#22853, @kriscon-db)
- [Tracking / UI] Add Portkey as a supported AI Gateway provider (#22830, @sairavuri-sudo)
- [Tracking] Cache successful basic-auth credential checks to eliminate per-request PBKDF2 (#22817, @PattaraS)
- [] Add `last_updated_at` field to model catalog entries (#22838, @copilot-swe-agent)
- [Evaluation] Enable third-party scorer registration in OSS MLflow (#22634, @smoorjani)
- [] Add platform-side telemetry fields to Gateway invocation and budget events (#22557, @PattaraS)
- [Tracing] Auto-start SQL warehouse before V4/V5 MLflow tracing calls (#22798, @artjen)
- [Tracing] Migrate Claude Code tracing to TypeScript-based plugin for simpler installation and runtime handling. (#22338, @B-Step62)
- [Tracing] Add TypeScript Qwen Code tracing via Stop hook (#22411, @kriscon-db)
- [] Support `model_kwargs` in DeepEval scorers for LLM parameter control (#22494, @debu-sinha)
- [] Add Application Default Credentials auth mode for Vertex AI gateway (#22754, @harupy)
- [UI] Reorganize Settings into section-based routing with sub-sidebar navigation (#22743, @serena-ruan)
- [Tracing] Add TypeScript Codex CLI tracing via `notify` hook (#22410, @kriscon-db)
- [Tracing / UI] Add Share button to trace detail view (#22608, @alkispoly-db)
- [Docs / Models] Add `mlflow.diffusers` flavor for diffusion model LoRA adapters (#22253, @Rasaboun)
- [Tracing] Add size limit for trace attachments (#22575, @kriscon-db)
- [Tracing] Add `GeminiCliTranslator` for Gemini CLI OTLP span type mapping (#22409, @kriscon-db)
- [Tracing] Add JSON OTLP encoding support for trace ingestion (#22408, @kriscon-db)
- [Tracing] Extract `service.name` from OTLP resource attributes for usage telemetry (#22407, @kriscon-db)
- [Tracing] Add MLflow tracing spans to guardrail execution in gateway (#22581, @TomeHirata)
- [Tracing / UI] Add rendering size guards for large media content (#22574, @kriscon-db)
- [Tracking] #21037 Add presigned upload URL endpoint for S3 artifact uploads (#21039, @henishborad)
- [UI] Add Guardrails tab to endpoint editor (#22360, @TomeHirata)
- [UI] Add click-to-expand modal for trace attachment images (#22461, @kriscon-db)
- [UI] Enhance Gateway quick start cards with logos, multi-model options, and compact variant (#22513, @xq-yin)
- [Tracing] Add @mlflow/vercel for better Vercel AI SDK tracing in Databricks UC (#2210

_[Truncated at 4000 characters — full notes: https://github.com/mlflow/mlflow/releases/tag/v3.12.0]_

### v3.12.0rc0

- Date: 2026-04-28
- Version: v3.12.0rc0
- Original notes: https://github.com/mlflow/mlflow/releases/tag/v3.12.0rc0
- Permalink: https://whatsnew.fyi/product/mlflow/releases/v3.12.0rc0
- Labels: Pre-release

- **added** — Add automatic tracing for Claude Code, Codex, Qwen Code, and Gemini CLI as standalone TypeScript-based plugins installable as CLI binaries
- **added** — Add OpenClaw tracing plugin as a TypeScript-based plugin installable via npm with automatic tracing of OpenClaw sessions
- **added** — Add Guardrail base class and JudgeGuardrail implementation for AI Gateway guardrails with pre-LLM and post-LLM safety checks
- **added** — Add multimodal trace attachments support for images, audio, and files with click-to-expand image modals, inline rendering, audio playback, and size guards
- **added** — Add mlflow.diffusers flavor for saving and serving diffusion models and LoRA adapters

We're excited to announce MLflow 3.12.0rc0, which brings powerful new capabilities for agent developers:

**Major New Features**:

- **Automatic Tracing for more AI Coding Assistants**: First-class tracing for Claude Code, Codex, Qwen Code, and Gemini CLI, now distributed as standalone TypeScript-based plugins installable as CLI binaries. Drop one into your CLI of choice and capture every prompt, tool call, and turn out-of-the-box. (#22338, #22410, #22411, #22409, #22853, @B-Step62, @kriscon-db)

- **OpenClaw Tracing Plugin**: New TypeScript-based tracing plugin for OpenClaw. Install via npm and get full automatic tracing of OpenClaw sessions, no SDK changes required. Captures every prompt, tool call, and response across the session, allowing you to inspect agent behavior directly in the MLflow UI. (#22717)

- **AI Gateway Guardrails**: Add safety checks to any agent endpoint with the new `Guardrail` base class and `JudgeGuardrail` implementation. Configure pre-LLM and post-LLM guardrails directly from the Gateway endpoint editor, with full DB persistence, REST API, and gateway-level execution. Protect production agents from unsafe inputs and outputs without writing wrapper code. Tracing spans for guardrail execution are included. (#21964, #21960, #21962, #22306, #22360, #22577, #22581, #22767, @TomeHirata)

- **Multimodal Trace Attachments**: Trace agents that handle images, audio, and files, and view them directly in the MLflow UI. Click-to-expand image modals, inline rendering in span Details and Timeline views, audio playback, multi-part chat normalization for Gemini and OpenAI Responses API, and size guards for large media so the UI stays responsive. (#22465, #22461, #22451, #22462, #22466, #22460, #22449, #22450, #22574, #22575, #21783, @kriscon-db)

- **`mlflow.diffusers` Flavor**: New first-class flavor for diffusion models. Save and serve LoRA adapters for image-generation pipelines with the same `mlflow.<flavor>.log_model` ergonomics as PyTorch or Transformers. (#22253, @Rasaboun)

Stay tuned for the full release, which will include even more features and bug fixes.

To try out this release candidate, please run:

\`pip install mlflow==3.12.0rc0\`

### ts/v0.2.0-rc.1 — TypeScript SDK 0.2.0 RC1

- Date: 2026-04-13
- Version: ts/v0.2.0-rc.1
- Original notes: https://github.com/mlflow/mlflow/releases/tag/ts/v0.2.0-rc.1
- Permalink: https://whatsnew.fyi/product/mlflow/releases/ts-v0.2.0-rc.1
- Labels: Pre-release

- **added** — Release candidate for @mlflow/vercel TypeScript package version 0.2.0

Release candidate for `@mlflow/vercel` TypeScript package with version 0.2.0: https://github.com/mlflow/mlflow/pull/22105