LLM ORCHESTRATOR

Overview 01 Plan First 02 Route Intelligently 03 Execute & Learn

Prompt Optimization LayerTune prompts so cheaper models match production quality.Hybrid Model ArchitectureExpensive orchestrator delegates to cheap sub-agents. No wasted spend on simple tasks.Smart Model SelectionAuto-picks the best model per prompt, using expensive models only when needed.Model Limitation AwarenessKnows model limits — escalates or reroutes instead of hallucinating.

Granular Plan ModeGenerates detailed step-by-step plans when needed, and skips planning for simple tasks.Pre-Execution Token EstimationSee token usage and projected cost before build commands run.Token Budgeting SystemSet token budgets per session, task, or team — enforced before compute runs.Structured Plan OutputOutputs interactive plan.html — reviewable decisions, not markdown walls.

Smart Context ManagementOnly pass context relevant to the latest prompt — not the full session history.Context & Memory ControlAdd or remove context and memory to control what the model sees.

Interactive Think ModePause, redirect, or correct the agent mid-run — no full restart.Skill BuilderBuild skills with test and evaluation modes to benchmark and fine-tune behavior.CLI + Local Web IDEProvides a terminal CLI and localhost web IDE with embedded chat.AI Collaboration ModeShared agent sessions and handoffs — break out of solo AI silos.

Cross-User LearningIndexes resolved failures org-wide, like Stack Overflow for AI agent runs.Automatic Repository RulesGenerates repo rules from your codebase patterns on init.

How it works overview

Phase 01

Plan First

Analyze before you infer.

Overview

Every Devtor session starts with structured planning. Granular plan mode kicks in when a task needs it — producing step-by-step execution plans that reduce hallucinations and token waste. Pre-execution token estimation and token budgeting surface projected cost before any inference runs.

In depth

Plan-first mode scores task complexity, estimates tokens, and applies your budget limits. When planning is warranted, Devtor generates detailed step-by-step plans — output as structured plan.html for review. You approve, adjust context, or cancel — so expensive models are never invoked blindly on work a cheaper agent could handle.

What happens

1Parse the natural-language task and intent
2Detect whether granular planning is required
3Estimate tokens, cost, and budget headroom
4Generate plan.html or lightweight routing preview
5Present the plan for approval or adjustment

Terminal output

terminal

$devtor run "Refactor auth module"

Analyzing task complexity...

→ Complexity: medium · planning required

→ Estimated tokens: 2,400 · budget: OK

→ Plan written: plan.html

→ Est. cost: $0.006

Benefits

Granular plans when complexity demands them
Pre-execution token and cost estimates
Token budgets enforced before compute starts
Structured plan.html output for reviewable decisions

Related capabilities

Build, test, and benchmark custom agent behaviors.

Smart Context Management

Only pass what's relevant. Not the full conversation history.

Pre-Execution Token Estimation

Know the cost before you commit the compute.

Token Budgeting System

Set spend limits. Stay inside them.

Granular Plan Mode

Detailed step-by-step plans when the task demands them.

Context & Memory Control

Add and remove context to control AI output.

Structured Plan Output

Plans as interactive HTML, not markdown walls.

Automatic Repository Rules

Coding conventions extracted from your codebase.

NextRoute Intelligently

Ready to orchestrate?

Installation guide and CLI — coming soon.