This documentation is also published in plain markdown, so an AI agent can read it without parsing HTML. The MCP server of the thinger CLI adds its own guides for agents. This page lists both and when to use each.

Documentation files

File	URL	Content
llms.txt	https://docs.thinger.io/llms.txt	Index of every page in the llms.txt format: one line per page with its title, the URL of its .md and its description, grouped by section.
llms-full.txt	https://docs.thinger.io/llms-full.txt	Every page of llms.txt, complete, in one file. Each page starts with a Source: line that gives its URL.
A .md per page	https://docs.thinger.io/<page>.md	One page in plain markdown. The URL is the address of the page with .md at the end: this page is https://docs.thinger.io/ai/docs-for-agents.md.

The .md version keeps all the content of the page. Tabs become headings, steps become numbered lines in bold, notices become quotes, and the agent prompts become quotes introduced by Prompt for an AI agent. Links point to the .md version of other pages, so an agent can follow them.

Give the docs to an agent

• For one topic, give the agent the .md URL of the page, or paste its content. It is the smallest input and the most precise.
• To let the agent find its way, give it https://docs.thinger.io/llms.txt: it reads the index and opens only the pages it needs.
• To load everything at once, add https://docs.thinger.io/llms-full.txt as a source in tools that index documents (a project's knowledge, a RAG index). It is large: prefer llms.txt in a conversation.

Ask your AI agent

Read https://docs.thinger.io/llms.txt, find the page about Data Buckets and explain how to export the data of a bucket to CSV.

Guides of the MCP server

The MCP server of the CLI also serves guides written for agents, with the knowledge they need to work on your account: how the concepts of the platform relate, the code of products, the administration of a server and the Dashboard SDK. They come with the CLI, so they match the tools of its version.

Guide	Content
concepts	How devices, products, Data Buckets, properties, alarms, endpoints, file storages, tokens, projects and dashboards relate, and which data source fits each need.
product-code	The JavaScript a product runs.
server-admin	Administration of a server: hosts, domains, brands, user accounts, OAuth2 clients, identity providers and proxies.
dashboard-sdk, dashboard-sdk-types	The Dashboard SDK reference and its TypeScript declarations, of the SDK version the app installs (in the node_modules of the folder where the agent works), or else of the version that comes with the CLI.

The agent reads them in two ways:

• With the tool thinger_doc_read: without id it lists the guides, with id it reads one. The agent decides when to read them.
• As MCP resources (thinger://concepts, thinger://dashboard-sdk…), for clients that let you attach resources to a conversation.

The full list, with the URI of each resource, is in MCP tools.

Which one to use

You want the agent to…	Use
Work on your account through the MCP server	Nothing: the agent reads the guides of the MCP server with thinger_doc_read when it needs them.
Answer questions about Thinger.io, or follow a guide of this site	The .md page, or llms.txt to find it.
Write firmware, call the REST API or build an integration outside the MCP	The .md pages of Devices and the REST API.
Build a dashboard app	Both: the agent reads the dashboard-sdk guide while it works, and you can point it to Build dashboards with AI.

Related pages

MCP toolsEvery tool the Thinger.io MCP server offers to AI agents, with its parameters, and every resource type the agents can reach.Connect an AI agentInstall the thinger CLI, sign in, and connect Claude Code, Claude Desktop or any other MCP client to your Thinger.io account.