Built for Developers

Context Before
Code.

RiddleCode understands your codebase, plans the work, executes tasks, runs tools, verifies results, and ships production-ready code using the world's most advanced AI models.

500+ AI Models
Multi-Agent System
Codebase Intelligence
Enterprise Ready
Live Demo

Watch RiddleCode
Build Software Live.

RIDDLECODE
STATUS:ACTIVE
AGENTS RUNNING: 24
TASKS TODAY: 12,418
auto-cycle
Model Router
task:Research
Gemini 2.5 Pro· Google selected
$riddle build "Create SaaS landing page for Acme AI"
Live Preview
rendering
app/(landing)/page.tsx — riddlecode
page.tsx
Hero.tsx
package.json
0/17 lines
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
main
|0 err riddlecore
Ln 1, Col 1tsx
Agent Fleet
5 specialized · orchestrated live
Agent Research
ACTIVE
Agent Design
ACTIVE
Agent Backend
ACTIVE
Agent Frontend
ACTIVE
Agent QA
ACTIVE
Install Riddle v1.4.0

Get Running In
Under 60 Seconds.

One command. Any surface. RiddleCode installs into your terminal, editor, or cloud — and immediately starts mapping your codebase.

riddle://install
signed
CLI

Install the RiddleCode CLI

Run autonomous engineering tasks from any terminal. Works on macOS, Linux, WSL, and Windows.

~6.4 MB
no telemetry
signed
open-source
macOS · Linux · WSL
$ curl -fsSL https://riddlecode.dev/install.sh | bash
Windows PowerShell
PS> irm https://riddlecode.dev/install.ps1 | iex
Windows CMD
$ curl -fsSL https://riddlecode.dev/install.cmd -o install.cmd && install.cmd
The Real Problem

Software Engineering Is Broken
Because AI Lacks Context.

Every team using AI to build software is hitting the same wall. It’s not the model. It’s the system around it.

P-01 · Pain

Context Collapse

  • Forgets your architecture
  • Loses prior decisions
  • Rewrites working code
  • Loops on the same fix
RiddleCode
Your AI Shouldn’t Forget
Your Codebase.
Persistent Engineering Memory

RiddleCode maintains a living memory of your system — architecture, decisions, conventions — across every session.

P-02 · Pain

No Repository Understanding

  • Reads only a few files
  • Guesses at architecture
  • Misses dependencies
  • You re-explain the system, every time
RiddleCode
We Build A Live
Understanding Of Your Entire System.
System Context, Not File Context

Repo-wide graph of files, APIs, services, schemas, and call paths — indexed and continuously refreshed.

P-03 · Pain

Tool Chaos

  • Dev → AI → Search → Docs → Terminal → Logs → AI
  • Endless context switching
  • You become the integration layer
RiddleCode
One Request.
Full Execution.
The AI gathers evidence itself

Native tools — terminal, search, docs, GitHub, DB, logs — orchestrated automatically before a model is called.

P-04 · Pain

Model Selection Is Broken

  • GPT? Claude? Gemini? Grok? DeepSeek?
  • No one actually knows
  • Quality changes weekly
RiddleCode
Stop Choosing Models.
RiddleCode Routes Automatically.
Intelligent Model Router

Every task is routed to the best model for that step — planning, implementation, reasoning, verification.

P-05 · Pain

Massive Token Waste

  • AI reads thousands of lines
  • Burns tokens, then guesses
  • Compute scales, accuracy doesn’t
RiddleCode
Intelligence Before
Inference.
Tools execute first. AI reasons second.

Deterministic tooling collects evidence first — models only see what is relevant. Lower cost, higher signal.

P-06 · Pain

AI Stops At Code. You Still Ship.

  • Generates implementation
  • You test, debug, deploy alone
  • No verification, no accountability
RiddleCode
Stop Generating Code.
Start Delivering Outcomes.
Outcome-driven execution

Every change is compiled, tested, linted and validated before it reaches you. The unit of work is a shipped result.

The New Stack

This Isn’t An AI Coding Tool.
It’s A New Software Engineering Stack.

The old flow stops at code. RiddleCode adds the layers that actually ship software — context, evidence, routing, execution, and verification.

Today
Broken

The Old Way

prompt → model → code · no context, no execution, no verification

User
Prompt
Model
Code
no memory
no tools
no verify
RiddleCode
Shipped

The New Stack

intent → context → tools → routing → execution → verification → result

User
Intent
Repository Understanding
Architecture Understanding
Tool Intelligence
Evidence Collection
Model Routing
Execution
Verification
Shipped Result
live
+ memory
+ tools
+ verify
We compete at the orchestration layer — not the model layer.
aboveClaude CodeCursorCopilotWindsurf
Model Orchestration

500+ Models.
One Intelligent Router.

Always use the best model for the task.

Planning
Claude
Routed automatically
Implementation
GPT
Routed automatically
Research
Gemini
Routed automatically
Reasoning
DeepSeek
Routed automatically
Verification
Claude
Routed automatically
500+
AI Models Available
Live Benchmarks

Engineering Model
Leaderboard

Live rankings from real software engineering workflows. Compare the most effective models for coding, planning, debugging, architecture, agents, and autonomous execution.

All Models · This Week
Picks based on real-world execution · View usage stats

Claude Opus 4.7

by Anthropic
#1

The most capable model for complex planning and orchestration across multi-file refactors.

7d trend
+3.9 pts
98.0
SWE-bench — Architecture & Planning — 200K ctx

GPT 5.5

by OpenAI
#2

Remarkably detailed and consistent across implementation, refactoring, and code review modes.

7d trend
+5.0 pts
97.0
SWE-bench — Implementation & Refactoring — 400K ctx

Gemini 2.5 Pro

by Google DeepMind
#3

Industry-leading 1M+ context for research, documentation synthesis, and large repository sweeps.

7d trend
+6.5 pts
96.0
SWE-bench — Research & Documentation — 1M+ ctx

DeepSeek V3

by DeepSeek
#4

Enterprise-grade reasoning model built for analysis, math-heavy logic, and verification chains.

7d trend
+6.8 pts
95.0
SWE-bench — Reasoning & Analysis — 128K ctx
Reliability

Every Result Is
Verified.

No more silent regressions or random quality shifts. Every change passes a deterministic verification pipeline before it ever touches your repo.

riddle://verify
passed
Compile
$ riddle.compile
Builds the change in a sandboxed runtime
ok412ms
Test
$ riddle.test
Runs your test suite + generated coverage
ok1873ms
Lint
$ riddle.lint
Style, types, dead code, security rules
ok218ms
Validate
$ riddle.validate
Semantic diff against architecture rules
ok967ms
4 / 4 stages passed·3,470ms total·0 regressions
ready to merge ↗
Anthropic, OpenAI and others have publicly acknowledged quality regressions. RiddleCode treats reliability as a system property — not a model property.
Platform

Everything Needed To
Build Software

Codebase Intelligence

Maps files, dependencies, APIs, databases and architecture.

Terminal Execution

Runs commands directly.

Multi-Agent Workflows

Specialized agents collaborate automatically.

Context Compression

Massive repositories without context loss.

Tool Intelligence

Uses built-in tools before consuming AI tokens.

Model Routing

Best model selected automatically.

IDE Integrations

VS Code, JetBrains, Terminal.

Enterprise Security

Full auditability and controls.

Integrations

Available Everywhere
You Work

Terminal
VS Code
JetBrains
Desktop App
Web App
GitHub
Enterprise

Built For Engineering
Organizations.
Not Just Developers.

Individual AI tools were never designed for the realities of large teams — permissions, audit, memory, security, and architecture consistency at scale.

SOC 2 Type IIISO 27001GDPRZero data retention
Audit Logs
Every request, tool call, model invocation
Access Controls
SSO, RBAC, scoped permissions per repo
Private Deployments
VPC, on-prem, air-gapped options
Hardened Execution
Sandboxed tooling, redacted prompts
Usage Analytics
Per-team token, cost, and outcome telemetry
Architecture Memory
Org-wide decisions enforced across repos
Social Proof

The Future Of
Software Engineering

0+
Models
0.0M
Lines Analyzed
0+
Tasks Automated
0.0%
Execution Reliability
The Vision

AI Models Are Commodities.
Context Is The Moat.

The next generation of software development won't be powered by a single model. It will be powered by systems that understand code, orchestrate intelligence, execute work, and continuously improve.

That's RiddleCode.

Private Beta · Limited Spots

Join The Waitlist.
Build With Us Early.

5 quick questions. We use your answers to prioritize access and shape the product.

riddle://waitlist
Step 1 of 5
Work email
We'll send your position confirmation here.
We never share your email. Unsubscribe in one click.

Build Faster.
Ship Smarter.

Join the next generation of autonomous software engineering.