Routing

Documentation Index

Fetch the complete documentation index at: https://manifest.build/docs/llms.txt

Use this file to discover all available pages before exploring further.

​

What is routing?

Instead of sending every request to the same model, use Manifest to route your queries to different models and providers.

​

The 4 routing types

On-the-fly routing (complexity and task-specific) happens in under 2 ms with zero external calls and extra cost. Default and custom routing are setup beforehand.

​

Default

Every agent has a default model — the one Manifest falls back to when no other routing rule applies. You set it once in the dashboard Routing page and can swap it at any time without touching your code. All requests that don’t match a complexity tier, specificity, or custom header rule are sent here.

​

Complexity

Manifest scores each incoming prompt across 23 dimensions and assigns it to one of four tiers: simple, standard, complex, or reasoning. Each tier maps to a different model in the dashboard, so cheap queries go to cheap models and hard queries go to the best ones automatically. Scoring runs in under 2 ms with no external calls.

​

How scoring works

23 dimensions grouped into three categories. The same scoring pipeline feeds both the complexity tier and the specificity detector. Keyword-based (14) — Scans the prompt for patterns like “prove”, “write function”, “what is”, etc. Structural (5) — Analyzes token count, nesting depth, code-to-prose ratio, conditional logic, and constraint density. Contextual (4) — Considers expected output length, repetition requests, tool count, and conversation depth. Each dimension has a weight. The weighted sum maps to a tier via threshold boundaries. A confidence score (0–1) indicates how clearly the request fits its tier.

​

Session momentum

Manifest remembers the last 5 tier assignments (30-minute TTL). Short follow-up messages (“yes”, “do it”) inherit momentum from the conversation, so they don’t drop to a cheaper tier unnecessarily.

​

Tier overrides

Some signals force a minimum tier regardless of the score:

​

Task-specific

On top of complexity, Manifest detects what kind of task the request is about across nine categories (coding, web browsing, data analysis, image generation, video generation, social media, email, calendar, trading). When a category is detected, the request is routed to the model you’ve pinned for that category, regardless of the complexity tier. You can toggle each category on or off per agent in the dashboard.

​

Available tiers

​

How detection works

Two signals feed the detector:

1. Keyword dimensions. The same trie scan that feeds the complexity score also counts matches in category-specific dimensions (e.g. codeGeneration and technicalTerms count toward Coding, webBrowsing toward Web browsing, emailManagement toward Email).
2. Tool names. When a request includes tool definitions, names with known prefixes boost the matching category — browser_* / playwright_* → Web browsing, gmail_* / outlook_* → Email, gcal_* / calendly_* → Calendar, and so on.

If a category crosses the match threshold, it wins. Otherwise no specificity is assigned and the request routes on complexity alone.

​

Custom

Send an x-manifest-tier or x-manifest-specificity HTTP header from your client to force a specific tier or category on any request. Manifest uses the header value directly (confidence 1.0) and skips automatic scoring. This is useful for isolating agent subtasks, A/B testing models, or integrating with orchestration frameworks that classify requests themselves. See the full request headers reference for accepted values.