Two APIs. One Platform.

Raw inference with 22,250+ open-source models via Featherless AI, zero 429 rate limiting with unlimited queueing, model coalition fallback across 41 model mappings, and per-client concurrency slots. Intelligent inference with 904 pre-built specialized AI agents, tiered memory (core/recall/archival), knowledge graphs, workflow orchestration, multi-agent collaboration, guardrails, event sourcing, and 49 advanced features across 347 API endpoints.

Raw Inference — raw.openrout.ing

https://raw.openrout.ing

OpenAI-compatible API with model coalition, unlimited queue, and per-client concurrency. Direct passthrough to Featherless AI infrastructure with Redis-backed rate limiting. Supports streaming (SSE) and non-streaming inference. 22,250+ models including Qwen3-8B, Qwen2.5-3B-Instruct, Qwen3-Embedding-0.6B, DeepSeek-V4-Pro, DeepSeek-V4-Flash, DeepSeek-V3.2, and more. Zero 429 errors — RPM and concurrency queue indefinitely, only token budget returns 429. Model coalition automatically falls back to similar-spec models when upstream is busy. Per-client concurrency based on tier (4 slots for starter, up to unlimited for enterprise). Token usage tracked per client per day in Redis.

  • 22,250+ models (Qwen3, DeepSeek V4, Llama, Mistral, Phi, etc.)
  • Streaming (SSE) + non-streaming inference
  • Zero 429 (unlimited queue, Redis-backed)
  • Model coalition (41 fallback mappings, 3 retries per model)
  • Per-client concurrency slots (tier-based: 4 to unlimited)
  • Token usage tracking (per client, per day, per model)
  • Embeddings (Qwen3-Embedding-0.6B, 1024 dimensions)
  • Upstream: https://api.featherless.ai/v1
  • Response headers: X-Tier, X-Model-Used, X-Coalition, X-Queue-Wait, X-Request-ID

Intelligent Inference — api.openrout.ing

https://api.openrout.ing

Enriched LLM inference with 904 pre-built specialized AI agents across medical, financial, security, architecture, creative, and executive domains. Every request runs through a context enrichment pipeline that adds relevant memories and knowledge before LLM inference. Tiered memory system (core/recall/archival) with temporal search and consolidation. Knowledge graphs with entity-relation mapping and auto-build from text. Workflow engine with chaining, LLM-powered execution, and custom workflow creation. Multi-agent collaboration with task delegation, inter-team delegation, and A2A protocol. Skills marketplace with publish/install/rate. MCP server integration. Guardrails with rule-based content safety. Event sourcing with append-only log and replay. Episodic memory with entity tracking and temporal facts. 347 API endpoints across v1, v2, v4, and v5 APIs.

  • 904 pre-built AI agents (medical, financial, security, architecture, creative, executive)
  • Context enrichment pipeline (X-Enrichment-Ms, X-Enriched headers)
  • Tiered memory: core (CRUD), recall (search), archival (long-term)
  • Memory consolidation + temporal search + context building
  • Multi-agent collaboration (delegation, inter-team, A2A protocol)
  • Knowledge graphs (entities, relations, auto-build, search, stats)
  • Knowledge bases (documents, semantic query, CRUD)
  • Workflow engine (templates, chaining, LLM execution, custom workflows)
  • Skills marketplace (publish, install, rate, browse)
  • MCP server integration (register, list, invoke tools)
  • Guardrails (rule-based content safety, audit logs)
  • Event sourcing (append, query, replay, trim, stats)
  • Episodic memory (ingestion, entity queries, temporal facts, batch)
  • Durable agents (crash-resistant, session management, structured chat)
  • Batch processing (submission, DLQ, retry)
  • Feature flags (per-client toggle, bulk, reset)
  • WASM tools (sandboxed execution)
  • Persistent workflows (crash recovery, resume)
  • Auto-build (tools, skills, agents, workflows from prompts)
  • A/B testing (create, list, results)
  • Reputation system (leaderboard, per-agent scores)
  • Cost analysis + performance metrics + agent heatmap
  • Tree of Thought reasoning (quick + deep)
  • Facts & extraction (turn-based, batch, store)
  • Intent classification
  • Summarization (agent + messages)
  • Tool generation (from description, from URL)
  • Self-learning (agent self-evaluation)
  • Hubs (skills, tools, knowledge, workflows)
  • Scheduler (cron-style, cancel, run history)
  • Observability (metrics, errors, usage, snapshots, database, restate)
  • Pipelines (CRUD, execute, run history)
  • Plugins (search, register, invoke, categories, task-based)
  • Templates (CRUD, render, A/B variants, analytics)
  • Rate limits API (tiers, configure, check, increment, cleanup)
  • Crypto payments (invoices, status, webhooks)
  • Billing (Square, checkout, subscriptions, history, credit, plans)
  • Authentication (JWT, API keys, session tokens, registration)
  • Model catalog (22,250+ models, categories, featured, search)

347 API endpoints. 49 features. One platform.

Click any card to expand details.

Raw Inference Features (6)

⚡OpenAI-Compatible APIraw▼

Full OpenAI API compatibility — drop-in replacement for any client that supports the OpenAI Chat Completions format.

Streaming + non-streamingModel coalition fallbackPer-client concurrencyToken usage tracking

⚡22,000+ Modelsraw▼

Access to 22,250+ models via Featherless AI integration including Qwen3, DeepSeek V4, Llama, and more.

Model searchModel categoriesFeatured modelsModel capabilities

⚡Embeddingsraw▼

Generate embeddings for semantic search, RAG, and similarity matching.

Qwen3-Embedding-0.6BBatch processingCosine similarity

⚡Zero 429 Rate Limitingraw▼

Unlimited queue — RPM and concurrency queue indefinitely. Never get a 429. Only token budget returns 429.

Per-client slotsRedis-backed queueTier-based limits

⚡Model Coalitionraw▼

When upstream model is busy/429, automatically falls back to very similar spec models before giving up.

41 coalition mappingsAuto-retry (3x)Transparent fallback

⚡Streaming SSEraw▼

Server-Sent Events streaming for real-time token generation.

Chunked transferCache-Control: no-cacheConnection: keep-alive
Intelligent Inference Features (43)

🧠904 AI Agentsv1▼

Pre-built specialized agents across medical, financial, security, architecture, creative, and more. Each with domain-specific system prompts.

Agent chat with memoryAgent recommendationsAgent reputation scoresAgent heatmap

🧠Context Enrichmentv1▼

Automatic message enrichment before LLM inference — adds relevant context, memories, and knowledge.

X-Enrichment-Ms headerX-Enriched flagAgent count header

🧠Agent Memoryv1▼

Tiered memory system — core, recall, and archival layers with temporal search.

Core memory (CRUD)Recall memory (search)Archival memoryMemory consolidationTemporal searchContext building

🧠Multi-Agent Collaborationv1▼

Delegate tasks between agents, collaborate on complex problems, and chain workflows.

Task delegationInter-team delegationAgent collaborationWorkflow chaining

🧠Knowledge Basesv1▼

Create, query, and manage knowledge bases for RAG applications.

Document managementSemantic queryKnowledge base CRUD

🧠Knowledge Graphv5▼

Build and query knowledge graphs with entities, relations, and auto-build capabilities.

Entity managementRelation mappingAuto-build from textGraph searchGraph statistics

🧠Workflowsv1▼

Create, execute, and chain multi-step workflows with LLM integration.

Workflow templatesCustom workflowsWorkflow chainingLLM-powered executionExecution history

🧠Skills Systemv1▼

Build, search, and install skills from the marketplace.

Skill buildingSkill searchMarketplace publish/installSkill ratings

🧠MCP Server Integrationv1▼

Register and manage Model Context Protocol servers for tool integration.

MCP registrationMCP tool listingMCP tool invocation

🧠Guardrailsv5▼

Content safety rules with logging and health monitoring.

Rule creationRule managementContent checkingAudit logs

🧠Agent Spawningv5▼

Dynamically spawn new agents with specific capabilities.

Multi-spawnSpawn trackingCapability-based discovery

🧠Schedulerv5▼

Schedule recurring tasks and agent executions.

Cron-style schedulingTask cancellationRun historyManual tick

🧠Observabilityv5▼

Full system observability with metrics, errors, usage, and snapshots.

Database statsError trackingUsage metricsSystem snapshotsRestate monitoring

🧠Pipelinesv5▼

Create and execute data processing pipelines.

Pipeline CRUDPipeline executionRun history

🧠Pluginsv5▼

Discover, register, and invoke plugins for extended functionality.

Plugin searchPlugin registrationPlugin invocationCategory browsingTask-based discovery

🧠Templatesv5▼

Create and render prompt templates with variant support.

Template CRUDTemplate renderingA/B variant testingAnalytics

🧠Tool Generationv5▼

Auto-generate tools from descriptions or URLs.

Tool generationURL-based generationTool prefilteringContext-aware prefiltering

🧠Tree of Thoughtv5▼

Advanced reasoning with tree-of-thought decomposition.

Quick reasoningDeep reasoning

🧠Facts & Extractionv5▼

Extract structured facts from conversations and text.

Fact extractionTurn-based extractionBatch extractionStore and query

🧠Intent Classificationv5▼

Classify user intent for routing and automation.

Intent classification

🧠Summarizationv5▼

Summarize agent conversations and message histories.

Agent summarizationMessage summarization

🧠Event Sourcingv5▼

Append-only event log with replay and trimming.

Event appendEvent queryEvent replayEvent statsEvent trim

🧠Episodic Memoryv5▼

Episodic memory with entity tracking and temporal facts.

Episode ingestionEntity queriesTemporal factsBatch processing

🧠Knowledge Searchv5▼

Search across articles, books, and general knowledge.

Article searchBook searchKnowledge enrichment

🧠Self-Learningv5▼

Agents can evaluate and improve their own performance.

Self-evaluation

🧠Hubsv5▼

Centralized hubs for skills, tools, knowledge, and workflows.

Skills hubTools hubKnowledge hubWorkflows hubLive workflows

🧠Rate Limits APIv5▼

Programmatic rate limit management.

Tier configurationPer-key limitsUsage checkingCleanup

🧠A2A Protocolv2▼

Agent-to-Agent communication with task delegation.

A2A messagingA2A broadcastTask creationTask completionAgent discoveryAgent cards

🧠Durable Agentsv4▼

Crash-resistant agents with session management.

Durable chatSession historyStructured chatWorkflow chatSession status

🧠Batch Processingv4▼

Submit batch jobs for bulk inference.

Batch submissionDLQ managementRetry failed items

🧠Feature Flagsv4▼

Per-client feature flag management.

Feature toggleBulk updatesFeature resetFeature listing

🧠WASM Toolsv4▼

WebAssembly-based tool execution sandbox.

WASM registrationWASM invocationWASM listing

🧠Persistent Workflowsv4▼

Crash-resistant workflows with state recovery.

Workflow persistenceRun statusCrash recoveryResume runs

🧠Auto-Buildv2▼

Automatically build tools, skills, agents, and workflows from prompts.

Capability searchAuto-build from description

🧠Multi-Spawnv2▼

Spawn multiple agents in a single operation.

Batch spawningSpawn status tracking

🧠Marketplacev1▼

Publish and install workflows, skills, and agents.

Publish listingsInstall listingsRate listingsBrowse marketplace

🧠A/B Testingv1▼

Create and manage A/B tests for agents and prompts.

Test creationTest listingTest results

🧠Reputation Systemv1▼

Agent reputation leaderboard with per-agent scoring.

LeaderboardPer-agent reputation

🧠Cost Analysisv1▼

Track costs and performance across all agents.

Cost trackingPerformance metricsAgent heatmap

🧠Crypto Paymentsv5▼

Cryptocurrency payment integration for subscriptions.

Invoice creationInvoice statusWebhook handling

🧠Billing & Subscriptionsv1▼

Full billing system with Square integration.

Checkout flowSubscription managementPayment historyInvoice listingStore creditPlan management

🧠Authenticationv1▼

JWT-based auth with API key management.

Email/password loginClient registrationAPI key CRUDSession tokens

🧠Model Catalogv1▼

Browse, search, and filter 22,000+ models.

Category browsingFeatured modelsModel searchModel detail
View Full API Docs → Postman Collection