harness-mcp-v2

What it does

Harness MCP provides 11 consolidated tools spanning 215 resource types across the Harness.io platform: pipelines, deployments, feature flags, cloud costs, chaos engineering, security testing, and more. Rather than implementing hundreds of one-tool-per-endpoint services, it uses a registry-based dispatch system where agents call harness_list, harness_get, harness_create, and similar operations on any resource type. Includes 32 pre-built prompt templates for common workflows: end-to-end builds and deployments, pipeline debugging, DORA metrics review, vulnerability triage, cost optimization, access control audits, feature flag rollouts, and pull request reviews.

Who it's for

DevOps engineers and platform teams managing Harness.io deployments who need AI agents to query pipeline execution history, debug failed builds, manage feature flag rollouts, triage security violations, and optimize cloud costs across multiple projects without hardcoded environment variables.

Common use cases

• Query pipeline execution history and debug failed builds across multiple projects
• Review cloud cost trends and recommendations to optimize spending
• Plan and manage feature flag rollouts and rollback strategies
• Triage security vulnerabilities and compliance violations
• Retrieve DORA metrics, deployment frequency, and lead times across teams

Setup pitfalls

• A secret was detected in the source code—treat the repository as sensitive and ensure API keys are never committed
• CI tests are currently failing—verify the build status before deploying to production
• Account ID is auto-extracted from PAT and SAT tokens (pat.<accountId>... or sat.<accountId>...), but standard API keys require HARNESS_ACCOUNT_ID env var
• Requires filesystem read and write access—sandbox accordingly in restricted environments