
Build digital products for free!
WHAT IS OLLMEC?
WHAT MAKES OLLMEC UNIQUE
Ollmec is an autonomous product foundry that turns raw ideas into shipped, production-grade applications with zero context rot and near-zero cost.
Product Lifecycle vs. Code Snippets
Legacy AI tools are traditional text editors with a chat sidebar. Ollmec reimagines creation as a complete Product Development Environment, orchestrating the full lifecycle: Natural Language Blueprint ➔ System Architecture & ADRs ➔ Dependency-Linked Backlog ➔ Multi-Agent Build ➔ Verified Deployment.
Recursive Swarm + DAG Sub-Sub-Agents
Eliminates prompt collision and cross-talk. Macro Domain Leads isolate context windows to prevent contamination, while Micro DAG Sub-Sub-Agents execute atomic graph tasks in parallel with upstream contract streaming and automated feedback loops.
Mid-Task Cross-Model Switchboard
Never locked to one model. Heavy frontier reasoning engines design architecture and schemas, while ultra-fast, free/sub-cent models generate boilerplate, run linters, and execute test suites mid-task.
Fly Under the Radar of Paid Tiers
Thanks to lossless compaction, our dynamic capability framework, and smart relay mechanisms, Ollmec flies under the radar of paid tiers -- effectively allowing you to build products for free.
Self-Annealing Reliability Loops
Deterministic 3-layer architecture (Directives ➔ Orchestration ➔ Scripts). Autonomous repair loops intercept build/linter errors, diagnose root causes, and heal code before human sign-off.
WHO IS OLLMEC FOR?
Engineered for visionary creators, lean founders, and technical architects who want to build and ship real software without engineering bottlenecks.
Roadmap to Shipped Production
From zero-code creators to seasoned design leads. Turn high-level feature roadmaps, design tokens, and user stories into live web and mobile applications -- handling features, branding, and deployment.
Build for Practically Zero Cost
Brainstorm product concepts, generate actionable backlogs, and launch complete production MVPs at virtually zero API cost using smart switchboard routing and free-tier LLMs.
Day-One Model Benchmarking
Test new open-weight and proprietary model releases immediately. Plug in custom endpoints in settings, benchmark against the capability matrix, and mix-and-match LLMs across the fleet.
Deterministic Technical Control
Enforce strict Architectural Decision Records (ADRs), schema contracts, and self-healing test loops without babysitting unpredictable chat-sidebar copilots.
High-Velocity Incubation
Deliver client-ready applications in days rather than months by spinning up deterministic 3-tier agent swarms across isolated worktrees.
DESKTOP PDE ARCHITECTURE
A deterministic 5-layer execution engine that eliminates hallucination, context decay, and the limitations of conversational AI sidebars.
Product Suite
Synthesizes unstructured requirements into actionable functional specs and user stories.
Dual-Tier Orchestration
Macro Swarm Orchestrators partition functional domains with Dynamic Context Optimization; Micro DAG Engines execute recursive sub-sub-agents with dependency resolution.
Smart Switchboard
Dispatches tasks down prioritized LLM ranking chains auto-benchmarked via ArtificialAnalysis with instant failover.
Memory & State
Proprietary state compaction prevents memory drift across 100+ multi-agent turns.
Execution Harness
Autonomous test-and-repair loops intercept stack traces and heal bugs before user review.
INCUBATION vs. FREEFORM
An integrated desktop suite for structured product incubation alongside an unconstrained sandbox for rapid prototyping.
Systematic Incubation
Guided architectural pipeline taking raw ideas to production deployments with verified constraints.
Unconstrained Sandbox
Frictionless, open canvas for ad-hoc debugging, live whiteboard experimentation, and multi-model REPL.
- Multi-Model Terminal & ScratchpadsExecute Bash, Python, and Node with multi-LLM pairing.
- Ad-hoc Bug Fixing & Linter LoopsInstant in-place code repair without touching product blueprints.
- Live Whiteboard & Architecture CanvasVisual diagramming with bidirectional code-generation sync.
- One-Click Promotion to SpecsPromote working prototypes directly into Product Mode roadmaps.
- Workspace Branching & RollbackIsolated worktrees with deterministic zero-loss checkpointing.
CAPABILITY MATRIX & AUTO-FALLBACK
Powered by a dynamic capability matrix with ranked LLM chains for every agentic task -- automatically falling back down the leaderboard upon rate limits or provider downtime.
Capability Matrix
Maps each agentic capability to a prioritized, ranked leaderboard of models. Fully user-editable or auto-generated via ArtificialAnalysis benchmarks.
- Live benchmark sync via ArtificialAnalysis
- Granular skill tiers (reasoning, diffs, bash)
- Full manual ranking & custom model overrides
Ranked Chain Invocation
Ollmec automatically invokes the LLM at the top of the capability chain for that exact task, optimizing execution quality and throughput.
- Top-of-chain dispatch per DAG node
- Zero prompt distortion across provider APIs
- Dynamic token & latency balancing
Automated Failover
If the primary model hits a rate limit, quota ceiling, or API outage, Ollmec seamlessly falls back to the next ranked model in real-time.
- Instant 429 rate-limit interception
- Automated drop to secondary & tertiary models
- Zero context loss or broken execution state
3-TIER RECURSIVE SWARM & DAG FLEET
Ollmec decouples orchestration into two cooperative tiers: Macro Domain Swarms (SO/DCO) and Micro DAG Pipelines. When combined, they form a hierarchical recursive sub-sub-agent tree.
Lead Orchestrator
Master coordinator managing global product lifecycle, user intent validation, cross-domain synthesis, and background servers.
SO / DCO Subagents
Domain leads (Architecture, UI, Auth, QA) operating in isolated DCO context windows with difficulty-based model routing (1–10).
DAG Sub-Sub-Agents
Granular node execution workers resolving topological dependencies, writing concrete code, and piping upstream dependency context.
4 Dynamic Modes
Configurable runtime modes: Unified Single Agent, Pure DAG Pipeline, Flat Domain Swarm, and Hierarchical Nested Swarm.
AUTONOMOUS EXECUTION LOOPS
A deterministic 3-layer architecture that intercepts build failures and heals code before it surfaces.
Directives (SOPs)
Markdown-based specifications defining business constraints, rules, and acceptance criteria.
Orchestration
Agent decision loops, DAG step planning, and switchboard tool routing across models.
Execution Scripts
Deterministic scripts executing file I/O, compilers, test runners, and migrations in sandbox.
LOSSLESS PSEUDOCODE COMPACTION
Proprietary state compression eliminates token bloat and context window rot, enabling near-zero marginal API cost.
State Compression
Converts verbose logs and source files into dense semantic pseudocode and AST representations with 90% token reduction.
Zero Context Rot
Agents remain razor-sharp on turn 200. No memory degradation, zero hallucinations, and zero architectural drift.
Build for Free
Compacted prompts combined with free-tier model routing allow full apps to be built for practically zero API cost.
CONVENTIONAL AI vs. OLLMEC PDE
Why next-generation product incubation cannot happen inside conversational text editor sidebars.
| CONVENTIONAL AI IDE | OLLMEC PDE |
|---|---|
File-Level Edits Only Operates on isolated snippets without full-system architecture or roadmaps. | Full Product Lifecycle Orchestrates Blueprint ➔ Architecture & ADRs ➔ Backlog ➔ Multi-Agent Build ➔ Deployment. |
Single-Thread Bottleneck One assistant tries to handle planning, coding, and debugging at once. | 3-Tier Recursive Fleet Macro Domain Leads (SO/DCO) + Micro DAG Sub-Sub-Agents running in parallel. |
Single-Model Lock-in Trapped with one LLM vendor per session with zero task-specific routing. | Intelligent Switchboard Relay Mid-task LLM dispatching down capability chains with automated 429 failover. |
Severe Context Rot Degrades past turn 10–15, causing memory drift and hallucinations. | Lossless Pseudocode Compaction Proprietary state compression keeps agents razor-sharp past Turn 200+. |
Continuous Token Burn Burns expensive frontier tokens for trivial boilerplate and syntax edits. | Near-Zero Marginal Cost Free-tier and sub-cent model dispatching builds complete production apps for free. |
Build products at zero token cost.
Join the private alpha for the world's first native Product Development Environment.