Mcp Oci

mcp-oci

[CI] [License: MIT] [npm]

A Model Context Protocol server for Oracle Cloud Infrastructure (OCI). It gives an MCP-capable client (Claude Desktop, Claude Code, Cursor, Copilot, …) the ability to discover live OCI resources, map how they relate, and generate reproducible Terraform — with behaviour controlled entirely by flags.

Think of it as a Playwright-MCP for your cloud: instead of rebuilding infrastructure knowledge by hand, the model can ask "show all VCNs in the prod compartment" and "generate Terraform for this compartment" and get structured, secret-free answers.

Features

• Live discovery — compartments, regions, and any resource via OCI Resource Search.
• Terraform generation — faithful HCL for known types (VCN, subnet, instance, bucket, compartment) and annotated skeletons for the rest; whole-compartment modules with provider variables.
• Dependency graph — nodes/edges plus a suggested provisioning order (dependencies first).
• Secrets never reach the model — every payload is redacted before return.
• Security flags — access modes, compartment/region allowlists, provisioning gate, dry-run, and JSON audit logging (see below).
• Standard auth — OCI config file (~/.oci/config) or instance principals. No credentials stored by the server.

Security model

Tools

Discovery (read): list_compartments, list_regions, search_resources, list_compartment_resources, get_resource

Terraform (read): generate_terraform, generate_compartment_terraform, build_dependency_graph

Quickstart — add to your agent

Published on npm as @dockndevai/mcp-oci. No clone or build needed — your MCP client runs it on demand with npx. Start in read-only mode; see .env.example for every variable and docs/CLIENTS.md for the full per-client guide.

Claude Code (CLI)

claude mcp add oci -e OCI_PROFILE="DEFAULT" -e OCI_MODE="read-only" -- npx -y @dockndevai/mcp-oci

Claude Desktop · Cursor · Windsurf — same block in claude_desktop_config.json, .cursor/mcp.json, or ~/.codeium/windsurf/mcp_config.json:

{  "mcpServers": {    "oci": {      "command": "npx",      "args": [        "-y",        "@dockndevai/mcp-oci"      ],      "env": {        "OCI_PROFILE": "DEFAULT",        "OCI_MODE": "read-only"      }    }  }}

OpenAI Codex CLI — in ~/.codex/config.toml:

[mcp_servers.oci]command = "npx"args = ["-y", "@dockndevai/mcp-oci"]env = { OCI_PROFILE = "DEFAULT", OCI_MODE = "read-only" }

VS Code (GitHub Copilot, Agent mode) — in .vscode/mcp.json:

{  "servers": {    "oci": {      "type": "stdio",      "command": "npx",      "args": [        "-y",        "@dockndevai/mcp-oci"      ],      "env": {        "OCI_PROFILE": "DEFAULT",        "OCI_MODE": "read-only"      }    }  }}

Configure

Point it at a standard OCI config profile. For safety, use an IAM user/policy with read-only (inspect/read) permissions on the compartments you want the agent to see.

Example prompts

• "List all compartments, then show every resource in the prod compartment."
• "Generate Terraform for VCN ocid1.vcn.oc1..…."
• "Build a dependency graph for compartment … and tell me the provisioning order."

Run from source (development)

Prefer the published package above. To run from a clone:

npm installnpm run buildnode dist/index.js   # with the environment variables set

Develop

npm run dev        # watch modenpm test           # security policy + terraform generation + graph + redactionnpm run typecheck

Roadmap

• terraform plan / apply execution behind read-write/admin + OCI_ALLOW_APPLY.
• More resource-type mappers (load balancers, databases, DRGs, IAM policies).
• Cross-environment drift comparison.

Publishing

This server ships a server.json for the official MCP registry and an mcpName for npm ownership validation. See PUBLISHING.md for publishing to npm and listing on the MCP registry, Smithery, Glama, Cursor, and PulseMCP.

License

MIT