Poolside Agent CLI logo
Coding agents

Use TokenRex with Poolside Agent CLI

Run Poolside's pool CLI against a custom OpenAI-compatible endpoint.

Custom-provider setup is documented in the linked official sources. A live TokenRex connection from this tool has not been tested. Screenshots illustrate the publisher's interface, not a verified TokenRex session.

Before you start

  • A current Poolside Agent CLI installation
  • A private TokenRex API key
  • An enabled streaming, tool-capable chat model

Setup methods

01

Poolside Agent CLI — custom provider

  1. 1

    Install Poolside's pool CLI using its official instructions. These settings configure the agent CLI, not a Poolside-hosted model account.

  2. 2

    Create a TokenRex API key and copy an exact enabled chat model ID from Models & Pricing. Replace the example ID with your selected model. Use verified context/output limits and a streaming, tool-capable model for agent tasks.

  3. 3

    Set TOKENREX_API_KEY privately in the environment that launches the app. Never share keys in screenshots, project files or exported configuration.

  4. 4

    Set POOLSIDE_API_KEY, POOLSIDE_STANDALONE_BASE_URL and POOLSIDE_STANDALONE_MODEL in the launch environment as below.

  5. 5

    Run the documented pool exec example with a short read-only question. The CLI reference also documents the custom endpoint override for interactive pool sessions.

  6. 6

    Send a short read-only question and confirm the exact model in TokenRex usage. Test streaming and tool calls separately before allowing file, terminal or connected-account actions.

Terminal commands
# TOKENREX_API_KEY must already be set privately
POOLSIDE_API_KEY="$TOKENREX_API_KEY" \
POOLSIDE_STANDALONE_BASE_URL="https://tokenrex.ai/api/public/v1" \
POOLSIDE_STANDALONE_MODEL="openai/gpt-6-astra" \
pool exec -p "Reply with a short greeting."

Setup reference

Troubleshooting

Authentication fails
Check that the saved TokenRex key is active and belongs in the tool's API-key field, not its subscription sign-in. Remove accidental spaces, check your wallet balance, and never share the key in screenshots or exported settings.
The model is missing or the request URL returns 404
Copy an exact enabled model ID from Models & Pricing. Follow this method's URL field: some tools expect the versioned base, others the root or full completion address. Check the documented API protocol before changing the path.
Chat works but streaming or tools fail
Test a short plain-text conversation first. Confirm the selected model and installed tool version support the required protocol and capabilities. Enable agent actions only after separately checking streaming and tool calls.

Verify the connection

Send a short message with the selected model, read its response, then confirm the exact model appears in your TokenRex usage. Test streaming and tool use separately before larger agent tasks.