Tools, MCP & Plugins

Tools are the capabilities — what an actor can actually do. Hermes exposes core tools and toolsets, speaks MCP (local stdio and remote HTTP), and supports plugins. This branch is about wiring capabilities deliberately and understanding the trust model, because the dangerous tools are the ones that write.

The rule that organizes everything: Skill = procedure, Tool = capability, Profile = actor, Cron = time, Memory = continuity, Messaging = interface. Tools are the verbs; keep read tools liberal and write tools gated.

Planned notes

• Core tools and the ~28 toolsets
• Tool schemas: how a capability is described to the model
• MCP basics: the protocol and why it matters
• Local stdio MCP servers vs remote HTTP MCP
• MCP auth: OAuth 2.1 with PKCE for remote servers
• Plugins: extending Hermes beyond MCP
• The trust model: read tools vs write tools
• Tool filtering and least-privilege per profile
• Integrations: GitHub, Gmail, Calendar, filesystem, browser, VDP, own APIs
• The capability rule: Skill / Tool / Profile / Cron / Memory / Messaging

Core sources

• Hermes — Feature: Tools — https://hermes-agent.nousresearch.com/docs/user-guide/features/tools
• Hermes — Feature: MCP — https://hermes-agent.nousresearch.com/docs/user-guide/features/mcp
• Hermes — Feature: Plugins — https://hermes-agent.nousresearch.com/docs/user-guide/features/plugins
• Hermes — Security (tool trust model) — https://hermes-agent.nousresearch.com/docs/user-guide/security

⚠️ Verify the /features/ paths before citing — some may have moved.

Connects to: Skills & Procedural Memory · Hermes Architecture · Agent Security & Ops · Memory & Personal Knowledge