← Back to choco.tools

Pillar 03

Agentic

Documentation built for the Model Context Protocol. Your docs aren't just readable by AI — they're structured for it.

For developers

Your docs speak agent

AI coding assistants are only as good as the context they receive. Most documentation tools serve HTML pages optimized for humans. choco.tools serves structured content optimized for machines.

Every choco.tools page is available as a Model Context Protocol resource. Claude, GPT, and any MCP-compatible agent can query your docs, reference specific sections, and cite versioned content — with the same cryptographic verification humans get.

MCP capabilities

Content as resources

Every document is an MCP resource with a stable URI. Agents fetch specific pages by path or ID.

Structured schemas

Frontmatter, headings, code blocks, and metadata are parsed into structured JSON — not flat text.

Agent authentication

Agents authenticate via API keys with scoped permissions. Read-only, write, or admin access per workspace.

Workflow triggers

Agents can trigger choco.tools workflows — run quality checks, generate changelogs, draft docs from PRs.

For business

AI-powered documentation ops

Stop paying writers to check for typos and broken links. choco.tools's agentic workflows run quality checks automatically — on every commit, on every PR, on a schedule.

8 built-in workflows

Typo check, broken links, style guide, changelog generation, API docs sync, draft feature docs, translations, SEO audit.

Claude API integration

Generation workflows use Claude with prompt caching for cost-efficient, high-quality content creation.

Custom workflows

Define your own quality checks and generation pipelines. Trigger them from the dashboard or via API.

For enterprise

Agent governance built in

Enterprise AI adoption requires governance. choco.tools provides it at the documentation layer:

  • Scoped agent permissions: each agent gets an API key with explicit read/write scope per workspace.
  • Audit trail for agent actions: every agent-generated edit is logged with the agent's identity, the prompt that triggered it, and the resulting diff.
  • Human-in-the-loop review: agent-generated content enters a review queue before publication. Humans approve, agents execute.
  • Budget controls: per-workspace token usage limits for Claude API calls prevent runaway costs.

MCP server: choco.tools/api/mcp | Auth: API key or OAuth2 | Protocol: MCP v1.0

The future of documentation is agentic

Start buildingBack to: Unbreakable →
Agentic — choco.tools | choco.tools