Economy model A
Handles bounded, local work
Answer→Tools→Continue
A’s prefix keeps growingTool round-trips keep the model stable
Intelligent routing for long-horizon agents.
Intelligent routingfor long-horizon agents.
A long task need not rely on one model from start to finish. Match the agent to the task, then choose the right model for each stage—guided by past performance and the work ahead.
JevOfficial extension · One call to choose and assessCode, research, analysis, and automation. Keep your workflow and spend compute where it counts.
One task. The right model at each stage.THE GOALFrom an idea to a verified result.
STAGE 01 · BOUNDED WORK
Economy model, steady progress
Understand → Use tools → Verify
Reuse the prefix throughout the stageContext compaction · ReassessAssess past performance and the work ahead
↗ STAGE 02 · DEEPER REASONING
Primary model, take it further
Reason → Implement → Check the result
Build a new prefix, then keep reusing itIllustration · Reassessment can also keep the same model
Efficient · Compute where it countsReliable · Bounded failoverAdaptive · Decisions grounded in execution
A long task is not one model choice
Stay stable within a stage.
Choose again at natural boundaries.
A new user prompt is not the only boundary. When a long session compacts and rebuilds its context, HiRoute gets another natural chance to choose; ordinary tool continuations stay on the current model.
Economy modelPrimary modelSelect a decision point to inspect the handoff · Illustration
User sets the goal → Agent works autonomouslyOne user task can span multiple routing turns
User feedback → ContinueNew input creates another decision point
Hand off to primary B
Example: the next stage needs deeper reasoning
Answer→Tools→Continue
Build a new prefix on BSwitch after compaction, then reuse consistently
Continue with model B
Example: performance is competent
Answer→Tools→Continue
B’s prefix keeps growingReassessment does not force a switch
A decision point during autonomous work
After context compaction, choose again.
Compaction rebuilds a long session around a new prefix. HiRoute uses that moment to assess the prior stage and the work ahead; the client does not need to detect or report compaction separately.
At this point
Assess the prior stageAssess the work already performed by model A
Choose the next branchChoose the primary branch; model B takes over
Official HiRoute decision extensionJev. One decision.
Jev. One decision.
Look back. Choose what comes next.
Combine the current task with execution history. One upstream request chooses the next branch and can assess the previous model’s competence.
Deploy the Jev extension ↗Swipe to explore the full diagram
Can an economy model handle what comes next?
Jev estimates the probability that the task is simple. Only consider economy when it meets your threshold; complex work stays on primary.
Does prior performance support saving?
A low assessment of the previous stage blocks cost-cutting. A high score cannot override task complexity; without a valid score, complexity alone decides.
Competence blocks risky cost-cutting; complexity creates opportunities to save.
Rules policy in the official extension · Configurable thresholds · Jev is optionalEvidence for the next decision
See more than who was chosen.
See how the work went.
Review stage competence over time for models selected by the current plan—without typing internal model IDs.
Real product components · Illustrative data, not a model benchmark. Unrated, partial, and not-yet-covered turns remain explicit.
A score keeps its scope
A score belongs to a plan, revision, model, and execution stage; new turns are not silently covered by an older assessment.
Evidence can refine work plans
A main agent can query high- and low-scoring tasks, inspect evidence, and propose a narrower work plan.
Route the task. Then the compute.
Choose the agent for the task.
Then configure models and compute.
Designed for coding, research, deep analysis, and general automation—not around one hard-coded agent.
Choose an agent by purpose
A main agent discovers allowed work plans, delegates a goal, and retrieves the result. Delegation and model routing remain independent.
Configure models and compute
Compose model sources, reasoning profiles, and ordered candidates; use smart saving, fixed, or free-first routing.
Keep failover bounded
Candidates satisfy tool, image, and context requirements; exhaustion stops explicitly and delivery is not silently switched.
Agent ecosystem
Not tied to one agent.
Not limited to coding.
Integration status follows real support; available and planned paths stay visibly separate.
Available now
CodexClaude Code
Planned
QoderPiDeepSeek HarnessOpenCodeHermes AgentOpenClaw
Order depends on real integration and validation.
Your policy, one interface
Your decision policy.
One HiRoute interface.
Build a decision service with Jev, an LLM, or your own policy. The general API supports allowed-branch choice and optional competence scoring.
Decision API documentation →Download OpenAPI ↓
DECISION API / SINGLE REQUEST
POST /v1/decisions
{
"branch_id": "smart_saving_simple",
"assessment": { "score": 0.82, "partial": false }
}Start with HiRoute
Keep your workflow.
Rearrange the compute.
Stable 0.1.0 is available for macOS Desktop / Linux headless.
Linux headless · 0.1.0
x86_64 · 59.9 MB
curl -fsSL https://hiroute.ai/install.sh | shInstallation and startup →SHA256 b54b62e48d95…Questions
Frequently asked questions
Does model switching destroy the KV cache?
Different models do not share KV cache. HiRoute aligns model changes with natural rebuilds after context compaction when possible, while ordinary tool continuations stay on the same model. Actual hits still depend on provider, configuration, and cache lifetime.
Can the decision service still see history after compaction?
It receives HiRoute’s independent simplified Agent-turn history: user input, accepted output, tool names and coarse outcomes, and executed branches. It is not a full transcript and can be marked partial after eviction or restart.
Must I deploy Jev?
No. Built-in rules are enough to start; Jev is an optional official extension, and you can implement your own service against the same Decision API.
Is competence a success rate?
No. It represents competence for a specific execution stage and must be interpreted with plan, revision, assessment coverage, and partial-evidence state.