Obscurify.ai for agents

Give your agent 200+ models through one OpenAI-compatible API, paid for from a prepaid balance. The account is a random number, with no email.

Hand this to your agent

add and configure https://obscurify.ai/skill.md

Your agent reads the skill file, asks you for a key, sets itself up and tests it. Nothing gets paid for without you: you add the credit yourself.

What you need

• A key. Create an account, add credit on the account page, then click Show API Key. New accounts may get $1 of starter credit while this month's promo budget lasts.
• Base URL: https://obscurify.ai/v1
• A model. deepseek/deepseek-v4-flash is an inexpensive start. The rest are on the model list.

The setup below reads the key from your environment:

export OBSCURIFY_API_KEY="sk-obs1_..."

OpenClaw

openclaw plugins install clawhub:@obscurify/openclaw-provider
openclaw onboard --auth-choice obscurify-api-key --obscurify-api-key "$OBSCURIFY_API_KEY"
openclaw models set obscurify/deepseek/deepseek-v4-flash

Prefer plain config over a plugin? Full OpenClaw tutorial.

Hermes Agent

hermes plugins install obscurify-provider
hermes plugins enable obscurify-provider
echo "OBSCURIFY_API_KEY=$OBSCURIFY_API_KEY" >> ~/.hermes/.env
hermes -z "Reply with exactly: ok" --provider obscurify -m deepseek/deepseek-v4-flash

Prefer plain config over a plugin? Full Hermes Agent tutorial.

Agent Zero

Start it with the key:

docker run -d --name agent-zero -p 127.0.0.1:50080:80 -v a0_usr:/a0/usr \
  -e OTHER_API_KEY="$OBSCURIFY_API_KEY" agent0ai/agent-zero

Open http://127.0.0.1:50080, go to Settings, then Models, and edit the Default preset. Set both Main Model and Utility Model to:

• Provider: Other OpenAI compatible
• Model name: deepseek/deepseek-v4-flash
• API base URL (under Advanced Settings): https://obscurify.ai/v1

Leave the Embedding Model as it is: the default runs inside the container.

Cursor

Cursor's own-key setting covers a fixed list of providers, so add Obscurify.ai as an MCP server. Its models become tools the agent can call, alongside Cursor's own model. Put this in ~/.cursor/mcp.json:

{
  "mcpServers": {
    "obscurify": {
      "command": "npx",
      "args": ["-y", "@obscurify/mcp"],
      "env": { "OBSCURIFY_API_KEY": "${env:OBSCURIFY_API_KEY}" }
    }
  }
}

Claude Code

Claude Code runs on its own models, so this adds Obscurify.ai as an MCP server too. Its tools are ask, chat, second_opinion, list_models, generate_image and balance.

claude mcp add -s user obscurify -e OBSCURIFY_API_KEY="$OBSCURIFY_API_KEY" -- npx -y @obscurify/mcp

Anything else

Use the base URL, your key and a model ID in any tool that takes a custom OpenAI endpoint. The API docs and OpenAPI spec cover the rest.

What it costs

Each request is charged from your balance at the model's price. Agents use far more tokens than chat, because each step resends the system prompt, tools and history: small tasks in our tests averaged about 20,000 tokens. Pay As You Go stops at 1M tokens a month, Casual at 10M, and Pro has no limit. See pricing.

What stays private

Prompts and responses aren't stored on our servers, and text requests only use zero-data-retention routes, which rely on providers' published policies. Providers don't see who sent a prompt, but they do see its text, and that includes any files your agent reads and sends. Read the privacy policy.