MCP Jenkins has severe definition quality gaps. Only 2 tools present; both have minimal descriptions and critically incomplete parameter documentation. The `build_jenkins_job_with_name` tool accepts a single string parameter with no validation constraints, format specification, or usage guidance. The `check_jenkins_version` tool has an empty parameters object ({}) and no description of what version format is returned or how to interpret it. Neither tool documents output schemas, error conditions, or recovery paths. Parameter descriptions are entirely absent, LLMs cannot infer what 'name' should contain (job path? display name? ID?). No error handling guidance is present; tools provide only success messages. The descriptions, while present, are generic and under 100 characters, falling short of the 194-character baseline for production tools. This server would require significant work before being safe for agent use.
Parameter 'name' in build_jenkins_job_with_name lacks description and validation constraints. LLMs cannot determine whether to pass a job name, job path, job ID, or display name. No format specification (e.g., 'alphanumeric with hyphens', 'case-sensitive', etc.).
check_jenkins_version has empty parameters object but no description of output schema. What format is the version string? (e.g., '2.387.1', release notes URL, build metadata?). LLMs cannot plan downstream actions without knowing the return type.
No output schema documentation for either tool. Production baseline requires 100% of tools to have documented return types. Without schemas, LLMs cannot validate responses or extract fields for chaining.
Add comprehensive parameter descriptions to 'name' in build_jenkins_job_with_name: 'The Jenkins job name or path (e.g., my-job or folder/subfolder/job). Names are case-sensitive. Use list_jenkins_jobs() to discover available jobs if you only have a partial name.'
Document the output schema for build_jenkins_job_with_name: Return a structured object with fields: { success: boolean, job_name: string, build_id: string, build_url: string, queue_item_id: string }. Include the build URL so the LLM can offer it to the user.
Document the output schema for check_jenkins_version: Return { version: string (semver format), build_number: integer, build_date: ISO8601_string }. Clarify the version format and whether it includes prerelease identifiers.
Add error handling and recovery guidance to both tools. Example: 'If job not found, call list_jenkins_jobs() to see available jobs. If Jenkins is unreachable, verify JENKINS_URL environment variable and network connectivity. Retryable errors: network timeout, HTTP 5xx. Non-retryable: job not found, invalid credentials.'
Implement a list_jenkins_jobs tool with pagination (limit, offset/cursor) that returns available jobs with descriptions. This enables job discovery and prevents blind guessing.
Add a dry-run mode to build_jenkins_job_with_name or implement a confirmation_required annotation. Example: 'Invoke with dry_run=true to preview which job will be triggered without actually starting a build. Always dry-run first if unsure.'
Expand build_jenkins_job_with_name description: 'Triggers an immediate build of a Jenkins job on the server at JENKINS_URL. The build runs asynchronously; monitor progress via check_build_status() with the returned build_id. Note: This operation has side effects (executes pipelines, may deploy code, consume resources). Use with caution. Idempotency: Re-invoking with the same job_name will queue a new, independent build; it does not return the previous build.'
Score history
Overall score trend
↑ 2 points across a rubric change (v1 → v2)
34/100
Scored
Grade
Overall
Spec posture
Rubric
2026-09-22
F
34
2026-07-28+
v2
2026-03-09
F
32
-
v1
Tool descriptions are extremely minimal (42 and 35 chars respectively). Baseline for production is 194 chars. Descriptions lack WHEN to use the tool, what prerequisites exist (e.g., 'requires JENKINS_URL env var and valid credentials'), or what failures look like.
No error handling or recovery guidance. Both tools return only success messages (e.g., 'Job X was triggered successfully'). What if the job does not exist? What if Jenkins is unreachable? How should the LLM retry? No documentation.
build_jenkins_job_with_name is a destructive operation (triggers CI/CD builds with side effects) but has no confirmation mechanism, dry-run option, or risk annotation. The tool can be invoked by an LLM without human oversight.
Parameter 'name' in build_jenkins_job_with_name is ambiguous. Jenkins API accepts job names, job paths (folder/subfolder/jobname), and sometimes IDs. No disambiguation in description forces LLM guessing. Should support lookup by partial name (discovery tool missing).
No pagination, filtering, or discovery tools. A real Jenkins server has dozens or hundreds of jobs. The server provides no way for an LLM to list available jobs, search by pattern, or understand what can be triggered. LLM must guess valid job names.
build_jenkins_job_with_name
Add parameter constraints to 'name': Specify minimum length (1 char), maximum length (typical Jenkins limit ~255 chars), allowed characters (alphanumeric, hyphen, underscore, forward slash for paths), and case-sensitivity rule.
Implement per-parameter type validation and clear error messages. Example: If 'name' is an integer or null, return 'Invalid job name type. Expected string, got <actual_type>. Job names must be non-empty strings (1-255 characters, alphanumeric with hyphens and underscores allowed).'
Add a tool to retrieve build status and logs: get_build_status(job_name, build_id) returning { status: enum(pending|running|success|failure|aborted), duration_seconds: integer, result_url: string, log_excerpt: string }. This enables agents to monitor and report on builds.