What's new in Blazemeter MCP

This release adds Streamable HTTP as an MCP transport option, documents OpenTelemetry configuration, pins the MCP SDK below 2.x, and updates the Ubuntu base image used for Docker and Linux binary builds.

New features

Streamable HTTP transport

• Explicit MCP transport selection: stdio, http, and docker
• Resolution precedence: CLI --mcp [TRANSPORT] → BZM_MCP_TRANSPORT → stdio
• http maps to FastMCP streamable-http with configurable bind settings:
  • FASTMCP_HOST (default 127.0.0.1)
  • FASTMCP_PORT (default 8000)
  • FASTMCP_STREAMABLE_HTTP_PATH (default /mcp)
• docker preserves existing Docker/stdio compatibility (MCP_DOCKER=true)
• Runtime setup refactored (resolve_mcp_transport / build_runtime) with unit tests

OpenTelemetry documentation

• README section covering default telemetry behavior, disabling telemetry, and pointing at a custom OTLP collector
• Documents env vars and CLI flags (--no-telemetry, --otel-endpoint, --otel-headers)
• Clarifies reported traces/metrics and that credentials are not included

Dependencies

• Constrained mcp[cli] to >=1.27.0,<2.0.0 to avoid incompatible major upgrades
• Lockfile updated accordingly

Build and distribution

• Docker and GitHub Actions Linux binary builds moved from Ubuntu 25.10 → Ubuntu 26.04

Upgrade notes

• Default transport remains stdio — existing installs are unchanged
• To use Streamable HTTP: --mcp http or BZM_MCP_TRANSPORT=http
• Pinning mcp<2.0.0 is a safeguard; no action required unless you previously forced a 2.x MCP package
• Docker/image consumers: base image is now ubuntu:26.04

What's Changed

• Upgrade ubuntu image to latest lts by @diego-ferrand in #90
• Add otel documentation by @diego-ferrand in #91
• Mob 50409 connect over streamable http by @alejandroaires in #94
• Add upper bound in mcp dependency by @Baraujo25 in #95
• Upgrade Ubuntu image and enhance MCP transport configuration by @Baraujo25 in #96

Full Changelog: v1.3.0...v1.3.1

Highlights

This release adds cross-account search for tests and executions, plus optional OpenTelemetry tracing and metrics for MCP tool calls.

New features

Search actions for tests and executions

• tests tool — new actions:
  • search — search tests across an account with filters for name, workspace, project, cloud provider, locations, creator, duration, engine count, and virtual users
  • search_filter_values — list allowed values for search filters (workspace, project, locations, etc.)
• execution tool — new actions:
  • search — search executions (reports) with the same filter model, plus run-date time frames (latest, today, yesterday, last_7_days, last_30_days, or custom range)
  • search_filter_values — list allowed values for execution search filters
• Search results include direct BlazeMeter URLs, workspace/project context, and pagination via page_index

OpenTelemetry observability

• Optional tracing and metrics for every MCP tool call via a new telemetry module
• Per-tool spans with action name, duration, client info, and distributed trace context (traceparent / tracestate)
• Metrics: mcp.tool.calls (counter) and mcp.tool.duration (histogram)
• Defaults to the Perforce gRPC OTLP collector; fully overridable via environment variables
• New CLI flags:
  • --otel-endpoint — override OTEL_EXPORTER_OTLP_ENDPOINT
  • --otel-headers — set OTLP headers (repeatable)
  • --no-telemetry — disable telemetry (OTEL_SDK_DISABLED=true)
• PyInstaller builds updated to bundle OpenTelemetry and gRPC dependencies

Dependencies

• Added: opentelemetry-api, opentelemetry-sdk, opentelemetry-exporter-otlp (and transitive gRPC/protobuf packages)

Upgrade notes

• Search and telemetry are opt-in / backward compatible — existing tool actions are unchanged
• To disable telemetry: --no-telemetry or OTEL_SDK_DISABLED=true
• To redirect telemetry: set OTEL_EXPORTER_OTLP_ENDPOINT (and optionally OTEL_EXPORTER_OTLP_PROTOCOL)
• Docker: telemetry env vars are documented in the Dockerfile; no config change required unless you want a custom collector

What's Changed

• Remediate mcp dependency CVEs by @Baraujo25 in #62
• Mob 45513 improve anomaly detection tool by @alejandroaires in #60
• Failure Criteria by @3dgiordano in #63
• Go from the deep link on terminal to setup wizard by @3dgiordano in #64
• Update dependencies (uv.lock) by @3dgiordano in #69
• Fix confirm mode in cli by @3dgiordano in #74
• Fix dist binary publication by @Baraujo25 in #75
• Fix confirm mode in cli by @Baraujo25 in #76
• Initial support for search actions by @3dgiordano in #68
• Otel support by @diego-ferrand in #84
• Version upgrade for release by @Baraujo25 in #86
• Release v1.3.0 by @Baraujo25 in #87

Full Changelog: v1.2.0...v1.3.0

What's new on the BlazeMeter MCP

Failure criteria

• Read failure criteria on tests with a structured domain model and human-friendly labels (KPIs, operators, and rule fields) so agents can explain thresholds without exposing raw API codes first.
• failure_criteria_meta — local, read-only catalog of KPIs, conditions, and rule fields (no BlazeMeter API call); use it to build valid rules before configuring.
• configure_failure_criteria — set or replace the full failure-criteria ruleset for a test (enable/disable, thresholds, sliding windows, stop-on-violation, and related options), with merge guidance when preserving existing rules.

Execution reporting and AI-assisted insights

• ai_analysis on the execution tool — start or poll BlazeMeter AI analysis for a finished (or running) execution; responses indicate when results are ready vs. still processing.
• read_anomalies_stats — anomaly statistics for an execution (status, counts, affected labels, per-anomaly context for LLMs), with clearer formatting and behavior when stats are unavailable for the account.
• read_summary — structured summary reports with normalized field names, metric explanations, and guidance aimed at better AI interpretation and troubleshooting (including pointers to relevant Skills where appropriate).
• read_errors and read_request_stats — richer, more consistent formatting and context for agents; improved handling when report data is not available (including archived executions).
• read_all_reports — simplified aggregation of summary, errors, and request stats in one call.

Reliability, security, and compliance

• Stronger validation and alignment with documented tool/action schemas across managers (reduces invalid or ambiguous MCP calls).
• Security-focused hardening: safer handling of paths and HTML in help, traceback sanitization, token and HTTP request constraints, skills and upload paths, and expanded automated tests (including Windows in CI where applicable).
• Dependency updates, including MCP stack CVE remediation and refreshed lockfile; python-multipart bump.
• Mend (SCA) scanning in CI

Developer and distribution experience

• Deeplink install — one-click style links on README.md to add the MCP in supported clients (e.g. Cursor, VS Code).
• Docker — image pull policy updated so runs use always pull for fresher images.
• Project hygiene — updated license, SECURITY.md (clearer reporting process), logo, and Node.js 24 in GitHub Actions.

Full Changelog: v1.1.1...v1.2.0

What's new on the BlazeMeter MCP:

• Minor changes related with improvements on help system.

What's new on the BlazeMeter MCP:

MCP Skills

• Modular, composable knowledge packages (instructions, scripts, troubleshooting guides) curated from official BlazeMeter documentation.
• Over 500 pages of documentation condensed into 15 categories and 88 resource files.
• Skills are loaded locally via Model Context Protocol (MCP) for fast, offline-ready, and secure AI-powered support.
• Fully transparent, open for community review and contribution, and aligned with the Agent Skills open standard.
  

Integrated Help System

• Enables users and AI agents to search the entire BlazeMeter documentation set for relevant information.
• Acts as a fallback when a specific Skill does not exist, ensuring users always have access to authoritative guidance.
• Seamlessly integrates with Skills for reference links and continuous knowledge updates.

Confirmation Mode

• It is possible to configure the confirmation mode for create, update, or delete operations.
• New argument --confirm allows you to configure the confirmation mode. Possible values: DELETE (Default) - Request confirmation on delete operations only; CUD - Request confirmation on create, update, and delete operations; DISABLE - Disable confirmation (warning: It allows an agent to perform any type of critical operation without human approval)

Delete support on Test

• The ability to delete tests is supported. By default, it will request confirmation from the user. This depends on the previously mentioned confirmation mode settings.

Billing and Costs Tool

• Calculate test costs based on test configuration for different test types.

Full Changelog: v1.0.1...v1.1

What's new on the BlazeMeter MCP:

• Execution support via uvx: It's now possible to run the project using uvx and a remote reference to our source code.
• Assigning an icon to a Windows executable file: Now executables in Windows have an icon.

Refer to the README file for further information and installation process.

Full Changelog: v1.0...v1.0.1

What's new on the BlazeMeter MCP:

• Performance Test Management: Create, configure, and manage performance tests with automated script uploads and asset management.
• Test Execution & Monitoring: Start tests, monitor execution status, and retrieve comprehensive reports including summary, errors, and request statistics.
• Workspace & Project Organization: Navigate through accounts, workspaces, and projects to organize your testing infrastructure.
• Load Configuration: Configure test parameters including concurrency, iterations, duration, ramp-up settings, and geographic distribution.
• Report Analysis: Access detailed execution reports, error analysis, and performance metrics for comprehensive test insights.
• Account & Permission Management: Manage multiple accounts and workspaces with proper AI consent controls and permission validation.

Refer to the README file for further information and installation process.

Full Changelog: https://github.com/Blazemeter/bzm-mcp/commits/v1.0