argocd-mcp

Exposes Argo CD application resources (status, sync state, health, revision) through the Model Context Protocol, allowing LLM clients to query live cluster state without direct kubectl access. Implements MCP resource endpoints that translate Argo CD API calls into structured JSON responses, enabling stateless queries of application metadata and deployment status across multiple clusters managed by a single Argo CD instance.

Implements MCP tool endpoints that trigger application sync operations (full sync, partial sync, refresh) against Argo CD, translating LLM tool calls into Argo CD API sync requests. Handles sync strategy configuration (auto-prune, self-heal, force), waits for sync completion, and returns operation status back to the LLM, enabling autonomous deployment workflows driven by LLM reasoning.

Provides MCP tools to generate diffs between desired (Git) and actual (cluster) application state, showing resource changes, manifest differences, and impact analysis. Implements handlers that call Argo CD's diff API, parse manifests, and format diffs for readability. Supports filtering by resource type and namespace.

Provides MCP tools to query Argo CD event logs and audit trails for applications, including sync operations, configuration changes, and user actions. Implements handlers that call Argo CD's event API, filter by timestamp/user/operation type, and format results for readability. Supports pagination and time-range filtering.

Provides MCP resources that expose real-time application health metrics (healthy/degraded/progressing), sync status (synced/out-of-sync/unknown), and resource-level health from Argo CD. Polls the Argo CD API to aggregate health conditions and surfaces them as queryable MCP resources, enabling LLMs to make decisions based on current cluster state without manual kubectl inspection.

Exposes MCP resources that query the git repository metadata associated with Argo CD applications, including current deployed revision, commit history, branch information, and git URL. Allows LLMs to inspect what code is currently deployed and retrieve commit details without direct git repository access, enabling context-aware deployment decisions and rollback reasoning.

Implements MCP resources that enumerate all applications across multiple Argo CD-managed clusters, with filtering by namespace, label selectors, and sync/health status. Aggregates application metadata from a single Argo CD instance managing multiple clusters, allowing LLMs to discover and reason about the entire deployment landscape without manual cluster enumeration.

Exposes MCP resources that retrieve the current application manifest, desired state from git, and actual state from the cluster, allowing LLMs to inspect what is deployed and compare against desired configuration. Provides structured access to Helm values, Kustomize overlays, and raw YAML without requiring LLMs to parse git repositories or kubectl output directly.

Implements MCP resources that retrieve event logs and pod logs for resources within an Argo CD application, providing visibility into deployment progress, errors, and runtime behavior. Aggregates Kubernetes events and container logs through Argo CD's API, allowing LLMs to diagnose deployment failures and understand application behavior without direct cluster access.

Exposes MCP resources that query Argo CD projects, RBAC policies, and access control rules, allowing LLMs to understand deployment permissions and constraints. Provides visibility into which applications a user/service account can access, what actions are permitted, and what repositories/clusters are allowed by policy.

Provides MCP tool endpoints that configure Argo CD notifications and webhooks, allowing LLM agents to register callbacks for deployment events (sync complete, health changed, error occurred). Enables event-driven automation where Argo CD notifies the LLM agent of state changes, triggering downstream actions without continuous polling.

Enables LLM clients to rollback applications to previous revisions or specific Git commits through MCP tools. Implements handlers that query application revision history, validate rollback targets, and trigger rollback operations via Argo CD API. Supports dry-run mode to preview rollback changes before execution.