Vault MCP Server

Implements the Model Context Protocol (MCP) server specification to expose encrypted vault operations through a standardized bidirectional message interface. Uses MCP's JSON-RPC 2.0 transport layer to handle tool definitions, resource schemas, and prompt templates, enabling any MCP-compatible client (Claude, custom agents, IDE extensions) to invoke vault operations without custom integration code. The server registers itself as a resource provider within the MCP ecosystem, allowing clients to discover and call vault methods through standard MCP tool-calling conventions.

Provides core vault operations for storing and retrieving encrypted data with integrated key derivation and management. Implements encryption at rest using industry-standard algorithms (likely AES-256-GCM or similar) with support for key rotation, versioning, and secure key storage. The server handles encryption/decryption transparently, accepting plaintext input and returning encrypted payloads on write, and accepting encrypted data on read with automatic decryption using managed keys. Key material is never exposed to clients; all cryptographic operations occur server-side.

Implements hierarchical path-based access control (PBAC) for vault resources, where permissions are granted at the path level (e.g., /secrets/prod/*, /secrets/dev/*). Clients discover available vault paths and their metadata through MCP resource endpoints, which return structured information about accessible vaults, their encryption status, and available operations. The server enforces access policies at request time, validating that the requesting client has permission to read, write, or delete at the requested path before executing operations.

Tracks and exposes vault operation metadata including creation timestamps, modification history, key versions used for encryption, and operation audit trails. The server maintains metadata for each stored secret (e.g., when it was created, which key version encrypted it, who last modified it) and provides MCP tools to query this metadata without decrypting the underlying data. Audit trails record all vault operations (read, write, delete) with timestamps and client identifiers, enabling compliance and forensic analysis.

Supports connecting multiple vault instances through MCP, enabling federation where a primary vault replicates encrypted data to secondary instances for high availability or geographic distribution. The server implements replication logic that synchronizes encrypted payloads and metadata across instances without exposing plaintext data. Clients can be configured to read from replicas for load balancing or failover, with the MCP protocol handling routing and consistency guarantees.

Supports atomic batch operations where multiple vault reads/writes are executed together with all-or-nothing semantics. The server implements transaction-like behavior where if any operation in a batch fails, all changes are rolled back. This is implemented through a batch request format where clients submit multiple operations in a single MCP call, and the server processes them sequentially with rollback capability if any operation fails.

Provides automated secret rotation where new versions of secrets are created and old versions are gradually phased out without disrupting client access. The server maintains multiple versions of each secret and supports gradual migration where clients can be configured to prefer newer versions while still accepting older versions during transition periods. Rotation is coordinated through MCP operations that create new versions, update client routing policies, and eventually retire old versions.

Exposes configuration options for encryption algorithms, key lengths, and cryptographic parameters through MCP tools. Clients can query supported algorithms (AES-256-GCM, ChaCha20-Poly1305, etc.), key derivation functions (PBKDF2, Argon2, etc.), and configure per-vault or per-secret encryption parameters. The server validates that requested algorithms meet security requirements and prevents downgrade attacks by enforcing minimum key strengths.

Stores and manages metadata (creation timestamp, last modified, owner, tags, TTL, rotation policy) alongside encrypted vault entries, enabling rich querying and lifecycle management. Implements tag-based filtering to allow clients to query vault entries by tags (e.g., 'production', 'api-keys', 'database-credentials') without decrypting data. Supports automatic expiration (TTL) and rotation policies that trigger re-encryption or deletion based on configured schedules.