NEW: CLAUDE OPUS 4.6 LIVE NOW

Claude Code, Codex, Gemini, Blackbox.
All Agents in One.

Blackbox unifies every major AI agent into one powerful platform. One subscription. One workflow. Unlimited possibilities. Join +30M builders on BLACKBOX.

BLACKBOX unifies every major AI agent into one powerful interface. One subscription. One workflow. Unlimited possibilities. Join +30M builders on BLACKBOX.

Get Started Now

Teams at fortune 500 companies that

depend on BLACKBOX.AI

BLACKBOX AI

signal-server

service

RateLimiter.java

RateLimitConfig.java

RateLimitMetrics.java

delivery

MessageDeliveryLoop.java

NoopDeliveryLoop.java

RedisMessageDelivery.java

challenges

ChallengeManager.java

ChallengeOption.java

RateLimitMetrics.java

RateLimiter.java

ChallengeOption.java

package service.limits;

import java.time.Instant;

import java.util.concurrent.*;

import java.util.concurrent.atomic.AtomicLong;

import java.util.concurrent.atomic.AtomicReference;

public class RateLimitMetrics {

allowedCount, blockedCount, resetCount

markReset(), getResetCount()

private final AtomicLong allowed = new AtomicLong();

private final AtomicLong blocked = new AtomicLong();

private final AtomicReference lastAllowed = new AtomicReference<>();

private final AtomicReference lastBlocked = new AtomicReference<>();

private final ConcurrentLinkedQueue samples = new ConcurrentLinkedQueue<>();

private final ScheduledExecutorService cleaner Executors.newSingleThreadScheduledExecutor();

public RateLimitMetrics() {

cleaner.scheduleAtFixedRate(this::trim, 1, 1, TimeUnit.SECONDS);

}

public void markAllowed() {

allowed.incrementAndGet();

lastAllowed.set(Instant.now());

samples.add(System.currentTimeMillis());

}

public void markBlocked() {

blocked.incrementAndGet();

lastBlocked.set(Instant.now());

samples.add(System.currentTimeMillis());

}

public long getAllowed() {

return allowed.get();

}

public long getBlocked() {

return blocked.get();

}

// inserted: approximate requests per second over last 60 seconds

public double requestsPerSecond() {

long now = System.currentTimeMillis();

long count = samples.stream().filter(t -> now - t <= 60_000).count();

return count / 60.0;

}

private void trim() {

long now = System.currentTimeMillis();

while (true) {

Long head = samples.peek();

if (head == null) break;

if (now - head > 60_000) samples.poll();

else break;

}

@Override

public String toString() {

return "allowed=" + allowed.get()

+ ", blocked=" + blocked.get()

+ ", rps=" + requestsPerSecond();

}

Task

Improve the rate limiting implementation.

Thought for 2 minutes.

I'll proceed with implementing all the improvements in a logical order starting with Enhanced Metrics and Observability, then Performance Optimizations, and finally Advanced Rate Limiting Features. Let me create a TODO.md file to track our progress:

TODO.md

BLACKBOX AI thinking...

Now let's start with Phase 1: Enhanced Metrics and

Observability. First, I’ll create a comprehensive metrics class for rate limiting:

RateLimitMetrics.java

Cancel

Type a message (@ to add files)...

Auto

BLACKBOX AI

signal-server

service

RateLimiter.java

RateLimitConfig.java

RateLimitMetrics.java

delivery

MessageDeliveryLoop.java

NoopDeliveryLoop.java

RedisMessageDelivery.java

challenges

ChallengeManager.java

ChallengeOption.java

RateLimitMetrics.java

RateLimiter.java

ChallengeOption.java

package service.limits;

import java.time.Instant;

import java.util.concurrent.*;

import java.util.concurrent.atomic.AtomicLong;

import java.util.concurrent.atomic.AtomicReference;

public class RateLimitMetrics {

allowedCount, blockedCount, resetCount

markReset(), getResetCount()

private final AtomicLong allowed = new AtomicLong();

private final AtomicLong blocked = new AtomicLong();

private final AtomicReference lastAllowed = new AtomicReference<>();

private final AtomicReference lastBlocked = new AtomicReference<>();

private final ConcurrentLinkedQueue samples = new ConcurrentLinkedQueue<>();

private final ScheduledExecutorService cleaner Executors.newSingleThreadScheduledExecutor();

public RateLimitMetrics() {

cleaner.scheduleAtFixedRate(this::trim, 1, 1, TimeUnit.SECONDS);

}

public void markAllowed() {

allowed.incrementAndGet();

lastAllowed.set(Instant.now());

samples.add(System.currentTimeMillis());

}

public void markBlocked() {

blocked.incrementAndGet();

lastBlocked.set(Instant.now());

samples.add(System.currentTimeMillis());

}

public long getAllowed() {

return allowed.get();

}

public long getBlocked() {

return blocked.get();

}

// inserted: approximate requests per second over last 60 seconds

public double requestsPerSecond() {

long now = System.currentTimeMillis();

long count = samples.stream().filter(t -> now - t <= 60_000).count();

return count / 60.0;

}

private void trim() {

long now = System.currentTimeMillis();

while (true) {

Long head = samples.peek();

if (head == null) break;

if (now - head > 60_000) samples.poll();

else break;

}

@Override

public String toString() {

return "allowed=" + allowed.get()

+ ", blocked=" + blocked.get()

+ ", rps=" + requestsPerSecond();

}

Task

Improve the rate limiting implementation.

Thought for 2 minutes.

TODO.md

BLACKBOX AI thinking...

Now let's start with Phase 1: Enhanced Metrics and

Observability. First, I’ll create a comprehensive metrics class for rate limiting:

RateLimitMetrics.java

Cancel

Type a message (@ to add files)...

Auto

BLACKBOX AI

signal-server

service

RateLimiter.java

RateLimitConfig.java

RateLimitMetrics.java

delivery

MessageDeliveryLoop.java

NoopDeliveryLoop.java

RedisMessageDelivery.java

challenges

ChallengeManager.java

ChallengeOption.java

RateLimitMetrics.java

RateLimiter.java

ChallengeOption.java

package service.limits;

import java.time.Instant;

import java.util.concurrent.*;

import java.util.concurrent.atomic.AtomicLong;

import java.util.concurrent.atomic.AtomicReference;

public class RateLimitMetrics {

allowedCount, blockedCount, resetCount

markReset(), getResetCount()

private final AtomicLong allowed = new AtomicLong();

private final AtomicLong blocked = new AtomicLong();

private final AtomicReference lastAllowed = new AtomicReference<>();

private final AtomicReference lastBlocked = new AtomicReference<>();

private final ConcurrentLinkedQueue samples = new ConcurrentLinkedQueue<>();

private final ScheduledExecutorService cleaner Executors.newSingleThreadScheduledExecutor();

public RateLimitMetrics() {

cleaner.scheduleAtFixedRate(this::trim, 1, 1, TimeUnit.SECONDS);

}

public void markAllowed() {

allowed.incrementAndGet();

lastAllowed.set(Instant.now());

samples.add(System.currentTimeMillis());

}

public void markBlocked() {

blocked.incrementAndGet();

lastBlocked.set(Instant.now());

samples.add(System.currentTimeMillis());

}

public long getAllowed() {

return allowed.get();

}

public long getBlocked() {

return blocked.get();

}

// inserted: approximate requests per second over last 60 seconds

public double requestsPerSecond() {

long now = System.currentTimeMillis();

long count = samples.stream().filter(t -> now - t <= 60_000).count();

return count / 60.0;

}

private void trim() {

long now = System.currentTimeMillis();

while (true) {

Long head = samples.peek();

if (head == null) break;

if (now - head > 60_000) samples.poll();

else break;

}

@Override

public String toString() {

return "allowed=" + allowed.get()

+ ", blocked=" + blocked.get()

+ ", rps=" + requestsPerSecond();

}

Task

Improve the rate limiting implementation.

Thought for 2 minutes.

TODO.md

BLACKBOX AI thinking...

Now let's start with Phase 1: Enhanced Metrics and

Observability. First, I’ll create a comprehensive metrics class for rate limiting:

RateLimitMetrics.java

Cancel

Type a message (@ to add files)...

Auto

BLACKBOX AI

signal-server

service

RateLimiter.java

RateLimitConfig.java

RateLimitMetrics.java

delivery

MessageDeliveryLoop.java

NoopDeliveryLoop.java

RedisMessageDelivery.java

challenges

ChallengeManager.java

ChallengeOption.java

RateLimitMetrics.java

RateLimiter.java

ChallengeOption.java

package service.limits;

import java.time.Instant;

import java.util.concurrent.*;

import java.util.concurrent.atomic.AtomicLong;

import java.util.concurrent.atomic.AtomicReference;

public class RateLimitMetrics {

allowedCount, blockedCount, resetCount

markReset(), getResetCount()

private final AtomicLong allowed = new AtomicLong();

private final AtomicLong blocked = new AtomicLong();

private final AtomicReference lastAllowed = new AtomicReference<>();

private final AtomicReference lastBlocked = new AtomicReference<>();

private final ConcurrentLinkedQueue samples = new ConcurrentLinkedQueue<>();

private final ScheduledExecutorService cleaner Executors.newSingleThreadScheduledExecutor();

public RateLimitMetrics() {

cleaner.scheduleAtFixedRate(this::trim, 1, 1, TimeUnit.SECONDS);

}

public void markAllowed() {

allowed.incrementAndGet();

lastAllowed.set(Instant.now());

samples.add(System.currentTimeMillis());

}

public void markBlocked() {

blocked.incrementAndGet();

lastBlocked.set(Instant.now());

samples.add(System.currentTimeMillis());

}

public long getAllowed() {

return allowed.get();

}

public long getBlocked() {

return blocked.get();

}

// inserted: approximate requests per second over last 60 seconds

public double requestsPerSecond() {

long now = System.currentTimeMillis();

long count = samples.stream().filter(t -> now - t <= 60_000).count();

return count / 60.0;

}

private void trim() {

long now = System.currentTimeMillis();

while (true) {

Long head = samples.peek();

if (head == null) break;

if (now - head > 60_000) samples.poll();

else break;

}

@Override

public String toString() {

return "allowed=" + allowed.get()

+ ", blocked=" + blocked.get()

+ ", rps=" + requestsPerSecond();

}

Task

Improve the rate limiting implementation.

Thought for 2 minutes.

TODO.md

BLACKBOX AI thinking...

Now let's start with Phase 1: Enhanced Metrics and

Observability. First, I’ll create a comprehensive metrics class for rate limiting:

RateLimitMetrics.java

Cancel

Type a message (@ to add files)...

Auto

State of the art performance

BLACKBOX Autonomous Agent

Run multiple autonomous agents in parallel locally
and remotely to execute your tasks securely.

Multiple layers of protection designed to meet the strictest security and compliance requirements.

Multi-Agent Remote Execution

Multi-agent remote execution with a chairman LLM to select the best execution for each task.

Get Started Now

Multi-Agent Execution

BLACKBOX dispatches the same task to multiple agents at once: BLACKBOX, Claude Code, Codex, Gemini.

Chairman LLM

Every task is implemented differently by each agent, and the chairman LLM selects the best implementation for users.

Long running tasks

Support for long-running tasks that require extended hours of execution in isolated sandbox environments where agents can install, implement, and test.

Monitor agents concurrently

One view to monitor all agents' implementations for individuals and teams collaborating on different projects.

BLACKBOX Agents on +35 IDEs

BLACKBOX Agents run tasks on the BLACKBOX IDE, VSCode, Jetbrains & more…

Get Started Now

Large code base context

BLACKBOX coding agent is optimized for both small and very large production-grade codebases that require extensive context.

Controllable Autonomy

Control the level of autonomy you want to grant the BLACKBOX coding agent while it executes your tasks.

Plan, Execute, Test

BLACKBOX agents are designed to implement tasks at a production level with detailed planning, high-precision execution, and a thorough testing phase.

Browser Agent

The BLACKBOX coding agent includes many built-in tools, including the browser agent, which provides autonomy in testing and iterating on its implementation.

BLACKBOX CLI

AI-powered development, right from your terminal.

Get Started Now

AI-Powered Code Generation

Generate APIs, scripts, components, and features directly from your terminal using simple natural-language prompts.

Smart Debugging & Fixes

Identify bugs, understand errors, and apply AI-generated fixes automatically without leaving the CLI.

End-to-End Project Automation

Set up projects, generate tests, configure CI/CD, and manage deployments — all from one CLI.

Works Across Any Stack

Use the CLI with Python, JavaScript, React, Node.js, databases, and more — no framework lock-in.