Skip to main content
Claude has two main ways to receive external context: the CLAUDE.md file (for Claude Code) and custom skills (for Claude Desktop). Both benefit from the Timely.ai llms.txt.

​
Claude Code

​
Inject context via CLAUDE.md

Claude Code reads the CLAUDE.md file at the project root before each session. Add an instruction so the model consults the Timely.ai llms.txt:
1

Download the llms.txt

At your project root, run:
curl https://docs.timelyai.com.br/llms.txt -o timely-llms.txt
2

Reference it in CLAUDE.md

Open or create the CLAUDE.md file and add:
## Timely.ai API

This project integrates with the Timely.ai API. The `timely-llms.txt`
file at the project root contains the full documentation for endpoints,
authentication, webhooks, and rate limits.

Always consult this file before suggesting code that interacts
with the Timely.ai API.
3

Verify loading

Start a Claude Code session with /init. The model will confirm it has read CLAUDE.md and will have the Timely.ai context available throughout the session.

​
Keep the file up to date

Add a script to your project to update the llms.txt when needed: 
#!/bin/bash
# scripts/update-timely-context.sh
curl -sL https://docs.timelyai.com.br/llms.txt -o ./timely-llms.txt
echo "Timely.ai context updated."

​
Claude Desktop — Custom Skills

You can create a skill that downloads the llms.txt and injects it as context into the system prompt before answering questions about the API.

​
Skill example

Create a skill file (.md format or as required by your Claude Desktop version): 
# Skill: Timely.ai Developer

## Description
Answers questions about the Timely.ai API using the official
documentation context.

## Instructions
1. When receiving a question about the Timely.ai API, fetch the context from:
   https://docs.timelyai.com.br/llms.txt
2. Use the returned content to ground your answer.
3. Cite the exact endpoint, the required parameters, and a code example.
4. If the question involves authentication, always mention the x-api-key header.

## Fixed context
API Base URL: https://api.timelyai.com.br
Documentation: https://docs.timelyai.com.br

​
MCP with context7

If you use the context7 MCP server configured in your environment, you can fetch Timely.ai documentation in real time during a session: 
Use context7 to fetch the Timely.ai documentation on
agent creation and show me the required fields.
Timely.ai indexing in context7 depends on the server being configured with access to the URL https://docs.timelyai.com.br. If the search returns empty, use the llms.txt directly as described above.
  1. Download timely-llms.txt with curl
  2. Add the reference in CLAUDE.md
  3. Commit the file to the repository
  4. Use /init to start sessions with context loaded
Open Claude Desktop, activate the Timely.ai Developer skill, and ask your question directly. The skill fetches the context on demand.
In Claude Code, paste the error and the endpoint:
I got a 422 calling POST /v1/contacts with the payload below.
What is wrong?
{ "name": "John", "phone": "47988001122" }
With the CLAUDE.md context loaded, Claude will compare against the expected schema and identify the problem.