Kendr combines an AI workspace, public model intelligence, local model support, hosted inference, compatible APIs and transparent usage records. Choose an outcome below to see what is available, where it runs and which current guide to open.
Each path points to a current Kendr surface. Use the labels to separate public research material from account, API and desktop capabilities.
Use the Kendr workspace to collect context, run research, preserve citations and turn the result into an artifact that can be checked later.
See the workspace 02 Developers · Account + API keyPoint a supported OpenAI- or Anthropic-format client at Kendr, select an available alias and retain settled usage and routing metadata.
Make the first request 03 AI and product leaders · PublicFilter sourced model profiles by provider, modality and availability, then inspect cost, context, capabilities, benchmarks and dated evidence.
Browse model evidence 04 Local-first users · Desktop + OllamaUse Kendr Desktop with models exposed by a connected Ollama runtime. The exact model, files and hardware determine what can run locally.
Review local model families 05 Operators and finance · Public + accountReview the fixed model-cost markup, live regional credit packages, non-expiring purchased credits and exact settlement fields.
Read the pricing policy 06 Technical evaluators · Public evidenceUse the public claims register to see the implementation evidence, scope and limitations behind routing, billing, security and reliability statements.
Inspect the evidence registerThese are product paths, not customer stories. Results depend on the source material, selected model, enabled route, account policy and the review applied to the output.
Frame a question, gather relevant sources, compare claims and preserve citations with the resulting artifact.
Compare model cost, context, capability, benchmark and scoped popularity data without treating every profile as a Kendr-hosted model.
Keep a supported OpenAI client, then change the base URL, API key and model alias. Test the request fields your application uses before production.
For Kendr managed routing responses, retain the public selected alias, decision metadata, fallback outcome, latency, version identifiers, usage and exact credit charge.
Use a compatible Ollama model from Kendr Desktop when the work should remain on the machine running the local runtime.
Quote from the current catalog, then use returned microcredit settlement and account billing activity as the record of completed usage.
Kendr has public knowledge, hosted execution and desktop-local surfaces. Choosing the correct boundary prevents a catalog page, a local model and a callable Cloud alias from being mistaken for the same thing.
Use individual model pages for research, the public model API for account-neutral hosted status, and GET /v1/models for aliases available to the signed-in API account.
The public catalog needs no key. Generating a response requires a funded Kendr account and a scoped API key kept outside browser code.
Keep a supported client. Change the base URL, API key and model alias. Persist kendr_usage, plus kendr_routing when a Kendr Route handled the request.
curl https://api.kendr.org/v1/chat/completions \
-H "Authorization: Bearer $KENDR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "kendr-intelligent",
"messages": [
{"role": "user", "content": "Summarize the tradeoffs."}
]
}'Inspect GET /api/public/models without a key, then use authenticated GET /v1/models for account availability.
Choose an exact model alias or a Kendr Route. The example uses kendr-intelligent. Use a stable idempotency key for safe retries of chargeable requests.
Store settled usage. For routed requests, retain the opaque receipt ID and sanitized route decision fields needed for support or audit correlation.
Public rates help estimate a request. The returned integer microcredit charge is the source of truth for completed model usage.
Exact model aliases and Kendr managed routing use the same fixed markup on configured provider model cost.
Read the ruleCredits from completed paid package purchases never expire. Promotional grants can carry their own expiry.
See wallet behaviorSuccessful model responses return settled usage and charged credit micros. Use those fields instead of reconstructing the bill.
Read usage fieldsKendr Routes expose a sanitized routing receipt with the selected alias, outcome, versions, latency, usage and cost.
Inspect evidenceSearch, media, knowledge-base, meeting and other non-model capabilities can have separate published unit rates. Do not infer those prices from the model markup. Open current pricing.
This hub describes current product paths. Volatile values and operational state belong in live APIs or explicitly dated evidence, not timeless marketing copy.
Account-neutral aliases, capability metadata, availability and public credit rates.
Open catalog JSONCurrent public package quote plus the published model markup and credit-expiry rules.
Open pricingA present-time website and public API database check, not an uptime history or SLA claim.
Check statusThe code, configuration or operational evidence that supports key public statements.
Review evidenceDirect answers to the availability, model, API and billing questions that determine the correct Kendr setup.
Kendr combines a web and desktop AI workspace with model discovery, hosted model access, compatible model APIs, routing receipts, local Ollama support and one account credit wallet. The right surface depends on whether you need public model research, hosted execution, local work or an API integration.
No. Individual model pages are labeled as Kendr-hosted, knowledge-only or local. Use the live public catalog for account-neutral hosted availability and authenticated GET /v1/models for the aliases available to an API account.
Yes, through Kendr Desktop and a connected compatible Ollama runtime. Local capability, speed and memory requirements depend on the exact model variant, quantization, hardware and runtime configuration. Sending the result into a hosted workflow changes the execution boundary.
Yes. Kendr documents supported Chat Completions and Responses routes and also supports documented Anthropic-format paths. A migration normally changes the base URL, key and model alias. Test the exact fields you rely on because Kendr does not claim compatibility with every upstream endpoint or extension.
For Kendr managed routing responses, kendr_routing is a sanitized record of the selected public alias, route decision category, outcome, versions, latency, settled usage and cost. It includes an opaque receipt ID for correlation while excluding provider credentials, internal router-model details and raw provider errors.
Kendr applies one fixed 5% markup to configured provider model cost for exact aliases and Kendr managed routing. Completed model responses contain the settled microcredit charge. Purchased credit packages never expire; promotional grants can have grant-specific expiration.
Start on the web for Kendr Cloud and public model research. Install Kendr Desktop when the work needs local files, a local Ollama runtime or desktop-only tools.