ReadVideo
← Back to ReadVideo

AI Agent · Setup guide

Let your assistant use ReadVideo.

Set up a delegated agent identity, then let your assistant submit recordings and retrieve transcripts. Start with the three steps below.

Set up an agent

  1. Create your agent

    Sign in to ReadVideo, open the Agents section, enter a name you recognize, and choose Create agent. If an existing agent shows no transcription access, enable it there.

    Open agent management

  2. Configure the credential

    Save the agent key when it is first shown. Add it to your assistant’s runtime environment or secret manager under this name:

    READVIDEO_AGENT_API_KEY

    Use the agent key, not an ordinary API key. Keep its value out of chat messages and source code.

  3. Give your assistant the instructions

    Once the credential is configured, copy these instructions into your assistant. Then give it an audio or video you are authorized to process.

    Read https://readvideo.ai/agent-skill and connect to ReadVideo through its aibot channel. READVIDEO_AGENT_API_KEY is configured in your runtime. Exchange it for a short-lived access token, then transcribe only the audio or video I explicitly provide and am authorized to process. Poll the task until it finishes and return the timestamped transcript, or explain the failure. Do not display credentials in chat.

    Your assistant needs to be able to read the setup document and make HTTPS requests. Pasting these instructions into a chat-only assistant does not add those capabilities.

For developers: token exchange and Agent Skill

The long-lived agent key is exchanged for a short-lived access_token. Send that token as a Bearer credential to the transcription endpoints. The agent needs the asr:user capability.

curl -s https://auth.xiaojinpro.com/oauth2/token \
  -d grant_type=urn:xjp:bot_api_key \
  -d client_id=readvideo \
  --data-urlencode api_key="$READVIDEO_AGENT_API_KEY"
Open the machine-readable Agent Skill

For help, email support@readvideo.ai.