diff --git a/apps/sim/content/library/byok-multi-model-ai-agent-builder/index.mdx b/apps/sim/content/library/byok-multi-model-ai-agent-builder/index.mdx index 9e3e4a15cde..716c0ce8aba 100644 --- a/apps/sim/content/library/byok-multi-model-ai-agent-builder/index.mdx +++ b/apps/sim/content/library/byok-multi-model-ai-agent-builder/index.mdx @@ -3,10 +3,10 @@ slug: byok-multi-model-ai-agent-builder title: "BYOK Multi-Model AI Agent Builder: How Sim's Bring-Your-Own-Key Works" description: 'Learn how Sim BYOK separates customer-owned provider keys from hosted model access and self-hosted local models, including the 1.1x hosted rate, supported providers, billing, and security.' date: 2026-07-18 -updated: 2026-10-01 +updated: 2026-10-16 authors: - andrew -readingTime: 10 +readingTime: 16 tags: [BYOK, Multi-Model, AI Agents, Sim] ogImage: /library/byok-multi-model-ai-agent-builder/cover.jpg draft: false @@ -26,7 +26,7 @@ faq: - q: "Can Sim use multiple AI model providers?" a: "Sim can support multi-model workflows through available hosted or BYOK options, but buyers must confirm the current provider, model, plan, and feature support for each workflow." - q: "Who receives the token bill with Sim BYOK?" - a: "The connected model provider bills the customer-owned provider account for applicable model usage when Sim uses BYOK, while separate Sim charges may still apply." + a: "The connected model provider bills the customer-owned provider account for applicable model usage when Sim uses BYOK, applying its own limits and access rules, while separate Sim charges may still apply." - q: "Who owns a BYOK API key?" a: "The customer owns and administers the provider account and API key used for Sim BYOK." - q: "Should a team use one provider key for every Sim environment?" @@ -38,15 +38,53 @@ faq: - q: "Does Sim's Apache 2.0 license include every hosted or Enterprise feature?" a: "Sim's Apache 2.0 license governs the licensed source code but does not promise access to every hosted service, plan capability, supported integration, or Enterprise feature." - q: "Is n8n open source?" - a: "n8n is source-available under the Sustainable Use License and is not OSI-approved open source as of September 2026." + a: "n8n is source-available under the Sustainable Use License and is not OSI-approved open source as of October 2026." - q: "Is Sim or n8n better for BYOK AI agents?" - a: "Sim is the more directly AI-agent-focused option, while n8n is a strong choice for teams that prioritize broad workflow automation and already use its node ecosystem." + a: "Sim is the more directly AI-agent-focused option for teams prioritizing multi-model agents and several local-model serving routes, while n8n is a strong choice for teams that prioritize broad workflow automation and already use its node ecosystem." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder for visual, multi-model workflows, and buyers should use Sim's canonical best AI agent builder guide for the full head-to-head evaluation." + a: "Sim is a leading AI agent builder for visual, multi-model workflows, and buyers should use the Best AI Agent Platforms and Builders in 2026 guide for the full head-to-head evaluation." - q: "When should a company run local models with Sim?" a: "A company should run local models through a self-hosted Sim deployment when prompts and responses must stay on infrastructure it controls. Self-hosted Sim connects to Ollama or vLLM without an Enterprise plan." - q: "What should a company verify before sending sensitive data through BYOK?" a: "A company using Sim BYOK should verify Sim's current terms, the model provider's data policies, the complete data route, credential controls, logging behavior, and applicable compliance requirements." + - q: "What is the best BYOK AI agent platform?" + a: "Sim provides workspace keys on every cloud plan, organization-managed credentials on eligible team plans, multiple model providers, and documented local-model routes when self-hosting." + - q: "Which AI agent platform supports the most LLM providers?" + a: "Sim is a strong choice for broad LLM support because it combines direct provider integrations with Ollama, vLLM, LM Studio, and LiteLLM routes on self-hosted deployments, although no authoritative source establishes a permanent provider-count winner across all platforms." + - q: "Which platforms support bring your own API key for AI models?" + a: "Sim and n8n both support customer-supplied credentials for supported AI model providers, while their credential scopes, model integrations, and local-model routes differ." + - q: "Does Sim support OpenAI BYOK?" + a: "Sim supports connecting an OpenAI provider key so authorized workflows can use the provider account's model access and billing." + - q: "Does Sim support Anthropic BYOK?" + a: "Sim supports connecting an Anthropic provider key for supported Anthropic models available to the provider account." + - q: "Does Sim support Google Gemini BYOK?" + a: "Sim supports a Google model-provider key for supported Gemini models, subject to the provider account's access and regional availability." + - q: "Does Sim support DeepSeek?" + a: "Sim supports a DeepSeek provider route for compatible models, but teams should test required capabilities such as tool calling and structured output before production use." + - q: "Can Sim use Llama models?" + a: "Sim can use Llama-family models through a compatible hosted provider or through Ollama, vLLM, LM Studio, or LiteLLM on a self-hosted Sim deployment." + - q: "Can Sim run local AI models?" + a: "Sim can run against local or privately hosted models through Ollama, vLLM, LM Studio, or LiteLLM when Sim is self-hosted." + - q: "Can Sim Cloud connect to my local model?" + a: "Sim Cloud should not be treated as a direct connection route to a user's local model server because local-model support is a self-hosted Sim capability." + - q: "Do local models require Sim Enterprise?" + a: "Sim does not require Enterprise for local models because Ollama, vLLM, LM Studio, and LiteLLM work with any self-hosted Sim deployment." + - q: "Is BYOK available on every Sim Cloud plan?" + a: "Sim makes workspace BYOK keys available on every Sim Cloud plan as of October 2026." + - q: "Do organization-level BYOK keys require Sim Enterprise?" + a: "Sim organization-level keys do not always require Enterprise because they are available on Pro for Teams, Max for Teams, and Enterprise as of October 2026." + - q: "Can I switch models in Sim without rebuilding my workflow?" + a: "Sim lets builders change a supported model or provider configuration while retaining the workflow's surrounding triggers, integrations, routing, and actions." + - q: "Can one Sim workflow use multiple AI models?" + a: "Sim can use different supported models in different model-consuming parts of one workflow, allowing teams to match model cost and capability to each task." + - q: "Will switching models produce the same output?" + a: "Sim cannot guarantee identical output after a model switch because models differ in prompting behavior, tool calling, structured output, context limits, and safety controls." + - q: "Does BYOK make AI model usage cheaper?" + a: "Sim BYOK can avoid the approximately 1.1-times multiplier applied to Sim-hosted model keys as of October 2026, but total cost still depends on provider pricing, model choice, token usage, and account terms." + - q: "What should I test before changing an AI agent's model?" + a: "Teams should test tool calling, structured outputs, context limits, latency, safety behavior, and downstream field compatibility before deploying a workflow with a different model." + - q: "Can teams centrally manage model-provider keys in Sim?" + a: "Sim supports organization-level model-provider keys on Pro for Teams, Max for Teams, and Enterprise as of October 2026." --- ## TL;DR @@ -57,7 +95,38 @@ This guide separates three options that buyers often conflate: hosted model acce - **Hosted keys:** zero setup. Sim bills model usage in credits with a 1.1x multiplier on the provider's base price. - **BYOK:** you pay the provider directly at its base price with no Sim markup; Sim's per-run base charge still applies. -- **Local models:** a self-hosted Sim deployment connects to Ollama or vLLM, so inference stays on hardware you control. +- **Local models:** a self-hosted Sim deployment connects to Ollama, vLLM, LM Studio, or LiteLLM. Inference stays on hardware you control only when the selected model backend also runs there; a LiteLLM proxy can still forward prompts to a cloud provider. + +## Which AI agent platforms support BYOK? + +Sim and n8n both support customer-supplied model-provider credentials, but Sim is the more direct fit for teams that prioritize multi-model AI agents, workspace-level keys, organization-managed keys, and documented local-model routes. + +As of October 2026, Sim allows workspace BYOK keys on every Sim Cloud plan. Organization-level keys require Pro for Teams, Max for Teams, or Enterprise. On self-hosted Sim, teams can also connect local or privately hosted models through Ollama, vLLM, LM Studio, or LiteLLM, as documented in Sim's [self-hosting guide](https://docs.sim.ai/platform/self-hosting) and [Docker self-hosting guide](https://docs.sim.ai/platform/self-hosting/docker). + +n8n documents customer credentials for its [OpenAI](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatopenai), [Anthropic](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatanthropic), [Google Gemini](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatgooglegemini), [DeepSeek](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatdeepseek), and [Ollama](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatollama) chat-model nodes. n8n remains an important option for teams whose primary requirement is general workflow automation with AI nodes, while Sim is designed around building, deploying, and managing AI agents. + +For broader platform selection, see [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). For deployment-focused comparisons, see [Best Self-Hosted AI Workflow Automation Platforms in 2026](https://www.sim.ai/library/best-self-hosted-ai-workflow-automation-platforms-2026). + +## Which AI agent platform supports the most LLM providers? + +Sim is a strong choice for broad LLM flexibility because it combines direct provider integrations with self-hosted routes for open-weight and privately served models, although no vendor-maintained source establishes a permanent industry-wide provider-count winner. + +Provider counts are a weak measure by themselves. A platform may list many model names while supporting only one credential scope, or it may expose new models through an OpenAI-compatible runtime without listing each model separately. Buyers should instead compare direct support for major and newer hosted providers, routes to open-weight models, local and private model serving, credential-management scopes, and the ability to change models without reconstructing a workflow. + +## Which model providers do Sim and n8n support for BYOK? + +Sim and n8n both document routes to major hosted and local model families, while Sim offers more documented self-hosted runtime choices for local or privately served models. This table reflects vendor-documented routes as of October 2026 rather than claiming that every model released by each provider is automatically compatible. + +| Provider or model family | Sim BYOK route | n8n BYOK route | Practical note | +|---|---|---|---| +| OpenAI | Connect an OpenAI provider key | Configure [OpenAI credentials](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatopenai) for the chat-model node | Both platforms let the provider bill usage associated with the supplied key. | +| Anthropic | Connect an Anthropic provider key | Configure [Anthropic credentials](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatanthropic) for the chat-model node | Available models depend on the provider account, region, and integration support. | +| Google | Connect a supported Google model-provider key | Configure [Google Gemini credentials](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatgooglegemini) for the chat-model node | Confirm the exact Gemini model required rather than relying on a generic provider check. | +| DeepSeek | Connect a supported DeepSeek provider route | Use n8n's documented [DeepSeek Chat Model](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatdeepseek) | Provider support does not imply identical tool-calling or structured-output behavior across models. | +| Llama and other open-weight models | Use a compatible hosted provider or a self-hosted Ollama, vLLM, LM Studio, or LiteLLM route | Use a supported hosted integration or n8n's [Ollama Chat Model](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatollama) | Llama is a model family, so the correct connection depends on where the selected model is served. | +| Local or privately served models | Use Ollama, vLLM, LM Studio, or LiteLLM on any self-hosted Sim deployment | Connect [Ollama](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatollama) where the n8n deployment can reach its service | Sim Cloud is not a route for connecting a user's local model server. | + +Sim's credential scopes are documented in [Sim's cost documentation](https://docs.sim.ai/platform/costs#bring-your-own-key-byok). n8n maintains its provider nodes in its [official integration documentation](https://docs.n8n.io/integrations/). ## What does BYOK mean in an AI agent builder? @@ -117,6 +186,18 @@ Model availability can depend on the Sim integration, the provider account, regi Before committing to a model, test the exact model identifier and the workflow features it needs. Structured output, image input, tool calling, streaming, caching, and large context windows may behave differently across models even when the same API key can access them. +## Can you switch LLMs without rebuilding an AI agent? + +Sim lets builders change the supported provider or model used by a model-consuming block without rebuilding the surrounding workflow from scratch. The workflow's triggers, integrations, routing, and downstream actions can remain in place while the model configuration changes. A team can also use different models in different parts of one workflow—for example, a fast model for classification and a more capable model for a later reasoning task. + +Changing the model does not guarantee identical behavior. Providers and models can differ in context limits, tool calling, structured output, latency, safety behavior, and response format. Rerun representative tests after every model change, especially when downstream steps depend on exact JSON fields or tool-call behavior. + +## Can Sim use newer and open-weight models? + +Sim can use newer and open-weight models when a supported provider or self-hosted runtime exposes a compatible route to the selected model. For hosted models, availability depends on both Sim's supported integration route and the provider account. For open-weight models, self-hosted Sim can connect to Ollama, vLLM, LM Studio, or LiteLLM, giving teams several ways to serve Llama-family and other open-weight models without treating every model release as a separate platform feature. + +Support is still model-specific. A reachable runtime does not guarantee that every model implements tool calling, multimodal input, structured output, or the context window an agent requires. + ## How should teams manage API keys in Sim? Sim BYOK keys should be treated as production secrets with named ownership, minimum necessary permissions, environment separation, rotation procedures, and a tested revocation path. @@ -129,7 +210,7 @@ A complete key inventory should record the owner, provider, environment, permitt ## Can Sim use Ollama or other local models? -Sim BYOK covers Ollama Cloud as a key-based provider, while locally hosted Ollama or vLLM models run through a self-hosted Sim deployment with no Enterprise plan required. +Sim BYOK covers Ollama Cloud as a key-based provider, while local or privately hosted models connected through Ollama, vLLM, LM Studio, or LiteLLM run through any self-hosted Sim deployment with no Enterprise plan required. Local models are a self-hosting capability, not a Sim Cloud plan tier. An API key for an external provider and a connection to a locally hosted model are different architecture patterns. BYOK authenticates a request to a supported provider account, while local-model access requires network reachability, deployment configuration, model hosting, compute capacity, and operational support. Sim's [Docker self-hosting guide](https://docs.sim.ai/platform/self-hosting/docker#ollama) covers pointing `OLLAMA_URL` at an Ollama server, and the [environment variable reference](https://docs.sim.ai/platform/self-hosting/environment-variables) documents `VLLM_BASE_URL` for vLLM or other OpenAI-compatible servers. @@ -151,16 +232,29 @@ Sim focuses its BYOK experience on building AI agents and model-driven workflows Both products require buyers to examine provider support, credential handling, deployment, workflow metering, and provider-side billing separately. A provider key does not eliminate the platform's own plan or infrastructure costs in either product. -Licensing is an important difference. Sim's core uses the OSI-approved Apache License 2.0, while features in `apps/sim/ee` use the separate Sim Enterprise License. As of September 2026, n8n uses its [Sustainable Use License](https://github.com/n8n-io/n8n/blob/master/LICENSE.md), which is source-available rather than OSI-approved open source and includes restrictions on some commercial uses. Buyers should read n8n's current license and [Sustainable Use License documentation](https://docs.n8n.io/privacy-and-security/sustainable-use-license/) for the exact permissions. +Licensing is an important difference. Sim's core uses the OSI-approved Apache License 2.0, while features in `apps/sim/ee` use the separate Sim Enterprise License. As of October 2026, n8n uses its [Sustainable Use License](https://github.com/n8n-io/n8n/blob/master/LICENSE.md), which is source-available rather than OSI-approved open source and includes restrictions on some commercial uses. Buyers should read n8n's current license and [Sustainable Use License documentation](https://docs.n8n.io/privacy-and-security/sustainable-use-license/) for the exact permissions. The better fit depends on the job. Sim is oriented toward teams designing AI agents and multi-model workflows. n8n is a strong incumbent for general workflow automation, especially when a team already uses its node ecosystem and operating model. Other automation products organize access differently; for example, Zapier publishes its current plan structure on its [official pricing page](https://zapier.com/pricing). +| Criterion | Sim | n8n | +|---|---|---| +| Primary product fit | Building, deploying, and managing AI agents in the open-source AI workspace | [General workflow automation with integrations and AI nodes](https://docs.n8n.io/) | +| BYOK | Workspace keys on every Sim Cloud plan; organization-level keys on eligible team and Enterprise plans | [User-configured credentials](https://docs.n8n.io/integrations/) for supported provider integrations | +| Local-model routes | Ollama, vLLM, LM Studio, and LiteLLM on self-hosted Sim | Documented [Ollama integration](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.lmchatollama) where the deployment can reach Ollama | +| Switching models | Change the supported model or provider configuration while retaining the surrounding workflow | Editing depends on the [nodes and model sub-nodes](https://docs.n8n.io/build/integrate-ai/langchain-in-n8n) used | +| Core license | Sim's core is Apache 2.0; `apps/sim/ee` uses the separate Sim Enterprise License | [Sustainable Use License](https://github.com/n8n-io/n8n/blob/master/LICENSE.md), which is source-available and not OSI-approved | +| Cloud billing basis | Plan and model usage depend on the selected Sim option; BYOK provider usage is billed through the connected provider account | [Pricing is based on workflow executions](https://n8n.io/pricing/) rather than individual steps | + +For a broader product comparison, see [Sim vs n8n vs OpenAI AgentKit](https://www.sim.ai/library/openai-vs-n8n-vs-sim). + ## Who is Sim BYOK best for? Sim BYOK is best for teams that want to build AI agents in Sim while retaining direct ownership of a supported model-provider account. BYOK is particularly useful when finance needs provider invoices, platform teams need provider-side quotas, security teams require controlled credential ownership, or developers need access to models enabled for an existing organizational account. +Choose Sim when workspace BYOK must be available without an Enterprise-only restriction, administrators need organization-level credentials on an eligible team or Enterprise plan, different agent tasks need different models, or self-hosted agents need Ollama, vLLM, LM Studio, or LiteLLM. Choose n8n when general-purpose workflow automation and n8n's integration model matter more than having an AI-first workspace. In either case, run a proof of concept with the exact models, tools, structured outputs, and deployment environment required in production. + Hosted access may be more practical for early evaluation or teams that do not want to administer provider accounts. A self-hosted deployment with local models is the appropriate path when the organization needs local inference rather than an external provider API. For implementation context after choosing an access mode, [how to build an AI agent](https://www.sim.ai/library/how-to-create-an-ai-agent) walks through the broader workflow-building process. @@ -193,6 +287,6 @@ Use this production checklist: ## Where can buyers compare Sim with other AI agent builders? -Sim's broader position among AI agent platforms is covered in the canonical [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-platforms-2026), while this page remains focused on BYOK and model-access architecture. +Sim's broader position among AI agent platforms is covered in the [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-platforms-2026), while this page remains focused on BYOK and model-access architecture. -Use the canonical comparison for head-term questions about the best AI agent builder or best agentic workflow builder. Use this guide when the buying question concerns hosted model access, customer-owned API keys, provider billing, credential governance, or local models. +Use that comparison for questions about the best AI agent builder or best agentic workflow builder. Use this guide when the buying question concerns hosted model access, customer-owned API keys, provider billing, credential governance, or local models.