OmniConvert for AI agents

AI agents can convert files with OmniConvert in four ways. All of them use the same credits: one credit per conversion, failed conversions are refunded, credits never expire. Files are processed in memory and never stored.

You get a token right after payment.

1. MCP server

A remote MCP server (Streamable HTTP) at https://www.omniconvert.cloud/api/mcp. Send your credit token in the Authorization header.

Claude Code:

claude mcp add --transport http omniconvert https://www.omniconvert.cloud/api/mcp \
  --header "Authorization: Bearer YOUR_TOKEN"

Or in a project's .mcp.json (other MCP clients use the same fields):

{
  "mcpServers": {
    "omniconvert": {
      "type": "http",
      "url": "https://www.omniconvert.cloud/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Tool	What it does
list_conversions	Search conversions by format or keyword. Free, no token needed.
convert_file	Convert content passed as text or base64. Text results come back as text, binary results as a file. One credit.
check_credits	Credits left on the token.

2. Agent Skill

A ready-made skill that teaches an agent when to use OmniConvert and how to call the API with curl. It reads the token from the OMNICONVERT_TOKEN environment variable.

Download the skill (ZIP), or read SKILL.md.
Claude Code: unzip it into ~/.claude/skills/, so the file ends up at ~/.claude/skills/omniconvert/SKILL.md.
Claude apps: upload the ZIP in the skill settings.

curl -sL https://www.omniconvert.cloud/skill/omniconvert.zip -o /tmp/omniconvert-skill.zip
unzip -o /tmp/omniconvert-skill.zip -d ~/.claude/skills/
export OMNICONVERT_TOKEN=YOUR_TOKEN

3. JSON API

Agents that make HTTP calls can send text or base64 and get JSON back, with no multipart upload or binary handling:

curl -s https://www.omniconvert.cloud/api/convert \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"conversionId": "csv-to-json", "text": "name,age\nAda,36"}'

# → { "filename": "input.json", "mimeType": "application/json", "size": 42,
#     "encoding": "utf8", "content": "[ ... ]", "method": "paid", "creditsRemaining": 59 }

encoding is utf8 for text formats and base64 for binary ones. Every conversion ID with its file extensions is in /api/conversions. The full reference, including multipart uploads, is in the API documentation.

4. OpenAPI and llms.txt

/openapi.json: an OpenAPI 3.1 spec, for example for custom GPT actions or API clients.
/llms.txt: a plain-text overview of the site and every conversion, for language models.

Safety

Agents should never send private keys, passwords or API keys. Key conversions accept public keys only and refuse private ones.
Treat the token like a password: anyone with it can use its credits. Use a separate credit pack per agent if you want to cap its spending.