Quickstart

Get Fishnet protecting your AI agent in under 5 minutes.

Prerequisites

Linux or macOS system
Rust 1.85+ (if building from source)
An AI agent using OpenAI or Anthropic APIs

Install Fishnet

Download and install

curl -fsSL https://github.com/iamyxsh/fishnet/releases/latest/download/install.sh | sh

The install script places the binary in ~/.local/bin/fishnet and creates the data directory at /var/lib/fishnet (Linux) or /Library/Application Support/Fishnet (macOS).

Initialize Fishnet

Run the first-time setup wizard:

fishnet init

You’ll be prompted for:

Master password: Encrypts your API keys (minimum 8 characters)
First API service: openai or anthropic
Credential name: A label for this key (e.g., my-openai-key)
API key: Your actual API key
Daily budget: Maximum spend in USD per day (default: $20)
Rate limit: Requests per minute (default: 60)

Example:

Master password: ********
Confirm master password: ********
First API service: openai
First credential name: my-agent-key
First API key: sk-proj-...
Daily LLM budget (USD) [20.0]: 10
Rate limit per minute [60]: 30

Your master password is never stored. Only a derived key is kept in your system keychain (optional).

Start the server

fishnet start

You should see:

[fishnet] config loaded from ~/.config/fishnet/fishnet.toml
[fishnet] listening on http://localhost:8473

Fishnet is now running as a local proxy.

Configure your agent

Point your AI agent to use Fishnet’s proxy instead of the real API:For OpenAI:

export OPENAI_BASE_URL=http://localhost:8473/proxy/openai

For Anthropic:

export ANTHROPIC_BASE_URL=http://localhost:8473/proxy/anthropic

Your agent will continue using its existing API key environment variable (OPENAI_API_KEY, ANTHROPIC_API_KEY), but Fishnet will intercept the requests and use the vault credential instead.

Test the setup

Send a test request through Fishnet:

curl
Python
Node.js

curl http://localhost:8473/proxy/openai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-anything" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

import os
from openai import OpenAI

os.environ["OPENAI_BASE_URL"] = "http://localhost:8473/proxy/openai"
client = OpenAI(api_key="sk-anything")  # Fishnet replaces this

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello"}]
)
print(response.choices[0].message.content)

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: 'sk-anything',  // Fishnet replaces this
  baseURL: 'http://localhost:8473/proxy/openai'
});

const response = await client.chat.completions.create({
  model: 'gpt-4o-mini',
  messages: [{ role: 'user', content: 'Hello' }]
});

console.log(response.choices[0].message.content);

The Authorization header can be any value — Fishnet overrides it with the vault credential.

Check status

View runtime metrics and today’s spend:

fishnet status

Output:

Status: running
PID: 12345
Uptime: 2h 15m
Active alerts: 0 (critical: 0, warning: 0)
Today's spend (USD):
  openai           $0.002340
Today's requests:
  openai           23

Verify with doctor

Run diagnostics to confirm everything is working:

fishnet doctor

Output:

Fishnet doctor report:
  [PASS] fishnet_running         process appears to be running
  [PASS] credentials_present     2 credential(s) in vault
  [PASS] openai_base_url         OPENAI_BASE_URL=http://localhost:8473/proxy/openai
  [PASS] anthropic_base_url      ANTHROPIC_BASE_URL=http://localhost:8473/proxy/anthropic
  [PASS] upstream_reachable      openai reachable (status 200)

Access the dashboard

Open http://localhost:8473 in your browser. The dashboard shows:

Real-time spend tracking
Active alerts
Request history
Credential management
Security policy configuration

What’s protected

Fishnet is now intercepting all AI requests and enforcing:

✅ Spend cap: Requests denied if daily budget exceeded
✅ Rate limiting: Max 30 requests per minute (from your init config)
✅ Credential isolation: Agent never sees real API keys
✅ Audit logging: Every request recorded with cryptographic proof
✅ Prompt drift detection: Alerts on unexpected system prompt changes

Next steps

Configure policies

Set spend limits, rate limits, and security policies

Explore security features

Learn about prompt drift, endpoint blocking, and ZK proofs

Set up alerts

Get notified via Discord/Slack when budgets or limits are hit

Integrate with your agent

Framework-specific guides for OpenClaw, LangChain, and CrewAI

Troubleshooting

Connection refused

If your agent can’t connect to Fishnet:

Check if Fishnet is running: fishnet status
Verify the port is correct (default: 8473)
Ensure OPENAI_BASE_URL / ANTHROPIC_BASE_URL are set correctly

Budget exceeded

If requests are denied with budget_exceeded:

Check today’s spend: fishnet status
Increase the budget: fishnet policy edit and update daily_budget_usd
Or wait until midnight UTC (daily limits reset)

Rate limit errors

If you see rate limit denials:

Check fishnet status for request counts
Increase the limit: fishnet policy edit and update rate_limit_per_minute
Or space out your requests

Get Started

Core Concepts

Security Features

Configuration

Operations

Integrations

Prerequisites

Install Fishnet

Check status

Verify with doctor

Access the dashboard

What’s protected

Next steps

Configure policies

Explore security features

Set up alerts

Integrate with your agent

Troubleshooting

Build docs developers (and LLMs) love

Get Started

Core Concepts

Security Features

Configuration

Operations

Integrations

​Prerequisites

​Install Fishnet

​Check status

​Verify with doctor

​Access the dashboard

​What’s protected

​Next steps

Configure policies

Explore security features

Set up alerts

Integrate with your agent

​Troubleshooting

Build docs developers (and LLMs) love

Prerequisites

Install Fishnet

Check status

Verify with doctor

Access the dashboard

What’s protected

Next steps

Troubleshooting