OpenCode

Use Fireworks AI models in OpenCode with the FireConnect CLI

FireConnect routes OpenCode through Fireworks AI models. See the FireConnect overview for install and CLI basics.

Change models: fireconnect opencode on --model <id>. See Models.

Prerequisites

• OpenCode installed
• A Fireworks API key (fw_...) or a Fire Pass key (fpk_...)
• FireConnect v0.9.1+ (see Install)

Enable Fireworks routing

fireconnect login
fireconnect opencode on

Restart OpenCode after enabling, then confirm routing:

fireconnect opencode status

Using Fire Pass

Use your fpk_... key during login or with --api-key:

fireconnect opencode on --api-key fpk_...

FireConnect detects Fire Pass keys and defaults OpenCode to kimi-fast-latest.

Default model

OpenCode routes a single default model (no opus/sonnet/haiku alias slots). The default is kimi-fast-latest, written to config as fireworks-ai/kimi-fast-latest.

Short model IDs like glm-5p2 are expanded to full Fireworks paths (for example, accounts/fireworks/models/glm-5p2).

What gets written

FireConnect merges a fireworks-ai provider block into ~/.config/opencode/opencode.json:

• An OpenAI-compatible adapter pointed at https://api.fireworks.ai/inference/v1
• A default model set to fireworks-ai/<model-id> (for example, fireworks-ai/glm-fast-latest)
• options.apiKey as a baked plaintext literal (file mode 0600)
• The preferred serverless catalog registered in the provider’s models for OpenCode’s /model picker

FireConnect snapshots your original opencode.json before the first change. The snapshot lives in ~/.fireconnect/opencode/. Running fireconnect opencode off restores the file byte-for-byte. OpenCode’s auth.json is never touched.

Browsing and picking models

fireconnect model list --search glm
fireconnect opencode on --model glm-5p2

fireconnect model list uses your stored key (keychain, config, or FIREWORKS_API_KEY). Fire Pass keys only list Fire Pass routers. FireConnect rejects --model firerouter with Fire Pass (fpk_...) on every harness; use an fw_... key.

FireRouter

fireconnect opencode on --model firerouter

Pass --anthropic-api-key sk-ant-... when your workspace does not have Anthropic BYOK provisioned server-side.

CLI reference

fireconnect opencode on              # Enable Fireworks routing
fireconnect opencode off             # Restore original config
fireconnect opencode status          # Check current provider and model
fireconnect opencode help            # Show harness-specific help

Run fireconnect opencode help for all options.

Switch models

fireconnect opencode on --model glm-5p2

Turn off Fireworks routing

fireconnect opencode off

This restores your previous opencode.json from the backup in ~/.fireconnect/opencode/.

Use a non-default config file

fireconnect opencode on --config-path /path/to/opencode.json

Fireworks on Microsoft Foundry

OpenCode supports Fireworks on Microsoft Foundry (CLI: --provider azure or on --azure). See the FireConnect overview and Microsoft Foundry integration guide for portal setup.

Configure and enable

export AZURE_API_KEY=<your-azure-api-key>

fireconnect configure \
  --provider azure \
  --base-url https://<resource>.services.ai.azure.com \
  --api-key $AZURE_API_KEY

fireconnect opencode on --model FW-GLM-5.2

One-off routing without changing global config:

fireconnect opencode on \
  --azure \
  --base-url https://<resource>.services.ai.azure.com \
  --model FW-MiniMax-M2.5

Pass your Foundry model with --model (for example, FW-GLM-5.2), not a Fireworks serverless short ID like glm-latest.

What gets written

FireConnect adds a fireworks-azure provider labeled Fireworks on Microsoft Foundry to opencode.json, pointed at your Foundry OpenAI-compatible endpoint (.../openai/v1). The default model reference becomes fireworks-azure/FW-GLM-5.2.

Use fireconnect model list only for browsing Fireworks serverless models on the direct gateway path. With Foundry, switch models with on --model FW-GLM-5.2.

Turn off Foundry routing

To switch back to the Fireworks gateway, change the global provider and re-enable:

fireconnect configure --provider fireworks
fireconnect opencode on

To remove FireConnect entirely and restore your original opencode.json:

fireconnect opencode off

See Turn off Foundry routing for details on global config behavior and uninstall.

Built-in provider connection

OpenCode also supports connecting to Fireworks directly without FireConnect:

1. Type /connect in OpenCode and search for fireworks.ai
2. Paste your Fireworks API key and press Enter
3. Type /models and select a model (for Fire Pass, choose a supported model such as GLM Fast Latest)

Source

FireConnect is open source: github.com/fw-ai/fireconnect