Fireworks · mantis-agent-sdk

Fast open-model serving with fine-grained model versions — note the long accounts/fireworks/models/... ids are normal.


endpoint	`https://api.fireworks.ai/inference/v1`
env var	`FIREWORKS_API_KEY`
get a key	app.fireworks.ai

Get an API key

Sign up or log in at app.fireworks.ai
Open Settings → API Keys
Click 'Create API Key'
Copy the key and store it securely
Export it as FIREWORKS_API_KEY

Create a key ↗ · Pricing ↗

New accounts get $1 free credit (10 req/min without a payment method).

Enable

export FIREWORKS_API_KEY=...        # shell profile — survives forever

or in the terminal — validates the key live before saving:

/enable fireworks

or pick any locked 🔒 Fireworks model in /models and paste the key inline.

Models

Starter menu (once enabled, /models fetches the provider's full live list):

accounts/fireworks/models/gpt-oss-120b
accounts/fireworks/models/deepseek-v3
accounts/fireworks/models/qwen3-235b-a22b

Switch anytime — /model gpt-oss-120b fuzzy-matches; context carries over.

SDK

from mantis_agent import MantisAgentOptions
options = MantisAgentOptions(
    model="accounts/fireworks/models/gpt-oss-120b",
    backend="https://api.fireworks.ai/inference/v1",     # key read from $FIREWORKS_API_KEY
)

Notes

Ids look like accounts/fireworks/models/gpt-oss-120b — paste them whole; /model oss fuzzy-matches fine.