# Enterprise Agentic Systems for Multi-Step Workflow Automation

> Fireworks enables real-time, reliable execution of multi-agent workflows across systems and APIs. Automate complex tasks, accelerate decisions, and scale enterprise operations with full control and governance.

## Open models built for your AI agents

Fireworks is the inference and training platform for AI agents. Genspark's research agent made 33% more tool calls than a state-of-the-art closed model, with 12% better quality, at 50% lower cost.

[Talk to our team](https://fireworks.ai/contact) · [Get started](https://app.fireworks.ai/signup)

Teams running agents on Fireworks: Genspark, Sentient, Notion, Cursor, Vercel

## How AI agents run on Fireworks

Start with inference on open models. Add training later if your tool-call patterns need it.

[Talk to our team](https://fireworks.ai/contact) · [Read the docs](https://docs.fireworks.ai)

- **Inference**: Start on serverless per token, then move to on-demand or Reserved Throughput as load grows; Sentient served 5.6M+ queries in a week with zero degradation. · [Deployment options](https://docs.fireworks.ai)
- **Training**: If an off-the-shelf model misses your tool-call patterns, fine-tune on multi-turn tool calls with SFT or RFT; Genspark went from POC to production in 4 weeks with RFT on Kimi K2. · [Fine-tuning docs](https://docs.fireworks.ai)
- **Ensembling**: Ensemble models through the agent frameworks that connect to Fireworks. · [Agent frameworks](https://docs.fireworks.ai)

## Open models with the tool calling your agents need

Why use Fireworks for AI agents?

- **Tool calling**: OpenAI-compatible tool calls, plus parallel calls on models that support them.
- **Responses API**: Connect MCP tools and cap how many tool calls one response can make.
- **Prompt caching**: Trilogy runs at ~1/5 the cost of proprietary systems, with 93.8% cache hits.
- **Lower latency**: Notion cut latency from about 2s to 350ms for 100+ million users.

## How to set up AI agents on Fireworks

How it works

[Read the docs](https://docs.fireworks.ai)

- **Step 1: Pick an agent model**: Pick from the recommended models for agents with tool use, such as GLM 5.2 or Kimi K2.6.
- **Step 2: Define your tools**: Describe each function in the OpenAI-compatible tools format and set whether and which tools the model can call.
- **Step 3: Constrain the output**: Turn on JSON mode or Grammar mode so every step returns output your code can parse.
- **Step 4: Connect your framework**: Plug Fireworks into a supported agent framework, or call MCP tools directly through the Responses API.
- **Step 5: Scale the deployment**: On a dedicated deployment, autoscale on tokens/s, requests/s or concurrent requests, and scale to zero when idle.

## How does Fireworks handle your agent data?

Security

Your agent sends prompts and tool results on every call, and you decide where that data sits and who holds the keys.

[Read the security docs](https://docs.fireworks.ai/guides/security_compliance/data_security)

- **Zero data retention by default**: Fireworks keeps no request data by default. If you train with Secure RFT, your reward functions stay with you.
- **SOC 2 Type II, HIPAA and ISO**: Fireworks covers SOC 2 Type II and HIPAA, plus ISO 27001, 27701 and 42001.
- **Encryption and your own keys**: Fireworks encrypts data with AES-256 and TLS 1.2+. You can bring your own storage bucket or encryption key.
- **Audit and access logs**: Enterprise accounts get audit and access logs for compliance reviews. Other account tiers don't include them.

## How to cut cost per task for AI agents

Guide

### Budget latency per call

Set a p95 latency budget for each call, then test every model you shortlist against it. If your budget depends on speculative decoding, plan for a dedicated deployment, the only place it runs. In a voice-to-action workflow, RADPAIR cut report turnaround from 15–20s to 2–5s.

### Make tool calls reliable before scaling

A failed tool call means a retry, and every retry adds to what the task costs. Reliability shows up on the bill. Use JSON or Grammar mode, and give each step only the tools it needs.

### Measure cost per task

Add up spend per completed task across every call. Innovative Solutions runs 4–10B tokens a month and moved 90% of its model spend to Fireworks in about 1.5 weeks. A full-parameter fine-tune runs on a dedicated deployment billed per GPU-second, so budget it apart from per-token traffic.

## Learn more about AI agents

Resources

[Browse the docs](https://docs.fireworks.ai)

- **[Open, frontier, and yours: LangChain Deep Agents on NVIDIA Nemotron 3 Ultra, running on Fireworks](https://fireworks.ai/blog/langchain-deep-agents-nvidia-nemotron-3-ultra)**: 9 JUL, 2026 · Announcement
- **[Agents don't fail on intelligence. They fail on execution.](https://fireworks.ai/blog/agent-execution-tax-browser-agent-benchmark)**: 20 MAY, 2026 · Learn
- **[Genspark’s Deep Research agent outperforms a frontier closed model in quality and tool calls using Fireworks RFT, achieving a 50% cost reduction](https://fireworks.ai/customers/genspark-deep-research-agent-rft)**: 31 OCT, 2025 · Genspark

## Frequently asked questions about AI agents

FAQ

[Read the docs](https://docs.fireworks.ai)

### Does Fireworks support function calling?

Yes. Function calling and tool calling are the same thing, and Fireworks runs both through its OpenAI-compatible API. Parallel calls depend on the model you pick.

### Which models work best for AI agents?

Start with a current serverless model with tool calling, such as Kimi K3 or GLM 5.3. Each model card lists its context length and whether it supports parallel tool calls.

### Which agent frameworks integrate with Fireworks?

The [agent framework integrations docs](https://docs.fireworks.ai/ecosystem/integrations/agent-frameworks) list every supported framework. If you skip a framework, step 4 above covers the Responses API route.

### Does Fireworks support structured outputs?

Yes. JSON mode holds output to a schema you supply, and Grammar mode holds it to a grammar you define.

### What does it cost to run AI agents?

On [serverless](https://docs.fireworks.ai/serverless/pricing), you pay per 1M tokens, with separate input, cached input and output rates for each model. Batch runs at 50% of serverless.

### Does Fireworks keep or log my agent data?

Not by default, and audit logs are Enterprise-only. The Security section above covers retention, encryption, keys and certifications.

## Build your first tool-calling agent

Run your agents on open models built for tool calling, then plan your deployment with our team.

[Get started](https://app.fireworks.ai/signup/) · [Talk to an expert](https://fireworks.ai/contact)
