Blocks.ai publishes this comparison and sells in this market. Every claim about another product links to a public source.
Kong AI Gateway vs LangSmith
Kong AI Gateway
Gateway that governs LLM, MCP, and agent-to-agent traffic
LangSmith
Platform for observing, evaluating, and deploying agents
Short answer
Kong AI Gateway is a gateway for LLM, MCP, and A2A traffic; LangChain calls LangSmith its platform for observing, evaluating, and deploying agents.Source 1, Source 2, Source 3 Kong proxies to agents registered as upstream URLs and can apply per-agent allow or deny lists; LangSmith Deployment runs agents in LangChain’s cloud (Plus or above) or, on Enterprise, self-hosted or hybrid.Source 4, Source 5, Source 6, Source 7
Where each one sits
Six layers of running AI agents at a company, and what each product’s own public sources say it covers.
These aren’t the same kind of product
- Where they overlap
- Both can authenticate calls to an agent, and both handle A2A and MCP: Kong proxies them, and LangSmith’s Agent Server serves them.Source 2, Source 4, Source 8, Source 9, Source 10
- Where they differ
- LangSmith builds agents without code, in Fleet, and runs agents on LangSmith Deployment.Source 5, Source 11 Kong proxies to agents registered as upstream URLs; agent-building tools in Kong AI Gateway aren’t publicly documented.Source 4
- Running both
- Kong says a LangGraph agent can send its model calls through Kong AI Gateway, and its walkthrough monitors the agent’s LangGraph Server with LangSmith.Source 12
Kong AI Gateway
- Build agents: Not publicly documented
- Host and run agents: Not publicly documented
- Agents across organizations: Not publicly documented
LangSmith
- Agents across organizations: Not publicly documented
At a glance
What each one is
Kong AI Gateway
Kong AI Gateway is one gateway for LLM, MCP, and agent-to-agent (A2A) traffic, with shared authentication, observability, and policy features.Source 1, Source 2 In 2.x, an agent is added as an AI Agent entity, which exposes it at a gateway endpoint and can carry policies.Source 4, Source 24
LangSmith
LangChain describes LangSmith as a framework-agnostic platform for observing, evaluating, and deploying agents.Source 3 LangSmith Deployment runs agents in production, and LangSmith Fleet creates and manages agents without code; LangChain positions Fleet for non-technical teams.Source 3, Source 5, Source 11 Agents hosted elsewhere can send traces to LangSmith.Source 6
The differences that matter
Who can call an agent
Kong AI GatewayThe gateway can require an API key or OpenID Connect login, such as Okta or Azure AD, then check a per-agent allow or deny list.Source 4, Source 17
LangSmithHosted agents use LangSmith API keys by default, or your own auth handlers; Fleet has per-agent Clone, Run, and Edit permissions.Source 8, Source 14
By default, Kong adds no credentials when it forwards a request to the agent.Source 4 Self-hosted Agent Servers have no default authentication.Source 8
Model and tool access
Kong AI GatewayKong gives one API to major LLM providers and can apply AI Policies to models and MCP servers; MCP bundling shows callers only authorized tools.Source 2, Source 20, Source 25
LangSmithFleet uses managed model tiers needing no provider key of your own; on Enterprise, admins can control who uses which tools and MCP servers.Source 18, Source 26
With Kong, you supply your own provider credentials.Source 2 LangChain says the model behind each Fleet tier may change over time.Source 26
What gets recorded
Kong AI GatewayA2A audit logs record task IDs, method calls, latencies, and errors, and A2A telemetry can flow to Konnect and OpenTelemetry.Source 4, Source 27
LangSmithEach deployment gets a tracing project, Fleet traces include tool calls and outputs, and LangSmith accepts traces from any OpenTelemetry-compatible app.Source 18, Source 28, Source 29
Konnect telemetry leaves out request and response bodies unless you opt in.Source 2 LangSmith keeps base traces 14 days and extended traces 180 days.Source 7
For security teams
What a security review asks, answered from each vendor’s public documentation.
Full comparison
18 criteria in five groups. Every cell links to its source, or says no public source answers it.
| Kong AI Gateway | LangSmith | |
|---|---|---|
| What it is | ||
| What it is and who it’s for | One gateway that governs LLM, MCP, and A2A traffic.Source 1, Source 2 All three run through one data plane, with shared authentication, observability, and policy features.Source 2 | LangChain calls it a framework-agnostic platform for observing, evaluating, and deploying agents.Source 3 It positions its no-code part, Fleet, for non-technical teams.Source 3, Source 11 |
| Maturity | Version 2.0 announced generally available on 1 September 2026; 2.2.0 released 30 September 2026.Source 20, Source 41 Konnect Catalog’s agent inventory is in beta.Source 13 | LangChain announced GA of LangGraph Platform (now Deployment) on 14 May 2025 and Agent Builder (now Fleet) on 13 January 2026.Source 15, Source 21, Source 22, Source 23 |
| Control | ||
| Agent registry and discovery | In 2.x, each AI Agent entity, A2A or plain HTTP, is scoped to one gateway instance.Source 2, Source 4 Konnect Catalog’s agent inventory is in beta.Source 13 | LangChain says Deployment manages agents in a centralized registry with versioning and rollbacks.Source 15, Source 42 Listing agents hosted elsewhere is not publicly documented. |
| Identity and access control | Can require an API key or OpenID Connect login, such as Okta or Azure AD, before routing, and check a per-agent allow or deny list.Source 4, Source 17 | Hosted agents use LangSmith API keys by default, or your auth handlers.Source 8 Fleet has per-agent Clone, Run, and Edit permissions.Source 14 RBAC is Enterprise.Source 43 |
| Ownership, policy, and revocation | Agent policies include input validation, logging, and rate limits.Source 4 Kong’s AI PII Sanitizer scrubs requests to AI models: a Plus add-on, included on Enterprise.Source 19, Source 44 | Fleet agent cards show the owner, sharing can be revoked at any time, and pausing an agent’s channels pauses it.Source 18, Source 23, Source 45 Admins can set spend limits.Source 46 |
| Audit log and observability | A2A audit logs record task IDs, method calls, latencies, and errors.Source 27 Konnect platform audit logs are Enterprise-only, sent to a webhook, and kept 7 days.Source 19, Source 36 | Enterprise audit logs of admin and configuration actions, kept up to 400 days.Source 37 Each deployment gets a tracing project.Source 28 |
| Connection | ||
| How agents connect | Each agent is an upstream URL the gateway proxies to.Source 4 In 2.x, data plane nodes you run forward allowed traffic and stay connected to Konnect.Source 2 | On self-hosted LangSmith with Deployment, agents run behind an ingress you configure.Source 47 Enterprise can reach LangSmith Cloud over AWS PrivateLink.Source 30 |
| Agents across organizations | Not publicly documented (checked 2 October 2026) | Agent Server agents can talk to other A2A-compatible agents.Source 9 An access model for another organization’s agents is not publicly documented. |
| Protocol support | Detects A2A over JSON-RPC and REST bindings.Source 4 Accepts four MCP revisions, 2025-03-26 to 2026-07-28 (the last from version 2.1).Source 48 | Agent Server supports A2A over JSON-RPC only, with documented gaps, and can expose agents as MCP tools.Source 9, Source 10 Fleet discovers tools from remote MCP servers.Source 16 |
| Frameworks, models, and clouds supported | Kong says it governs A2A traffic without changing how agents are built.Source 49 Agents that don’t use A2A can pass through as plain HTTP.Source 4 | Deployment runs Google ADK, Strands, CrewAI, and AutoGen agents via a wrapper or the LangGraph Functional API.Source 50 Agents hosted elsewhere can send traces.Source 6 |
| Operations | ||
| Deployment options and data residency | 2.x: a Konnect-managed control plane in a region you choose, data plane nodes you run.Source 2, Source 51 Kong also sells Kong-managed Dedicated Cloud and serverless gateways.Source 19, Source 52, Source 53 | Cloud on GCP (US, EU, APAC) or AWS (US).Source 6, Source 30 On Enterprise: BYOC (GA on AWS), self-hosted, or hybrid.Source 6, Source 7, Source 54 |
| Compliance attestations | From 2.2, FIPS mode uses only FIPS 140-3 approved algorithms; not submitted for NIST validation.Source 38 Kong lists ISO 27001, SOC 2, and PCI DSS for Kong Inc.Source 39 | LangChain’s Trust Center lists SOC 2 Type II, ISO 27001:2022, GDPR, and HIPAA, with a 2026 SOC 2 Type II report and ISO 27001 certificate.Source 40 |
| Support and SLA | Konnect targets 99.9% availability; Dedicated Cloud Gateways list a 99.99% SLA, serverless gateways none.Source 19 Enterprise support SLAs: 30 min to 2 hours.Source 19 | Base, Standard, and Premium support plans.Source 55 Reasonable commercial efforts toward 99.5% API uptime for SaaS, measured quarterly.Source 55 |
| Time and effort to get running | A quickstart script creates a Konnect control plane and a local Docker data plane.Source 25 To route A2A traffic, create an AI Agent entity and attach policies.Source 24 | Cloud deployment needs the Plus plan or above and an API key.Source 56 Self-hosting the Deployment control plane needs Kubernetes, self-hosted LangSmith, and Enterprise.Source 5 |
| Pricing model and public prices | Plus: from $25 a month plus usage; per control plane, $200 hybrid, $500 Dedicated Cloud, $25 serverless.Source 19 Enterprise: custom, billed annually.Source 19 | Developer: one free seat.Source 7 Plus: $39 per seat a month, with one free small serverless deployment.Source 7 Enterprise: contact sales.Source 7 Usage is metered in LSUs, $1 each.Source 7 |
| Building | ||
| Agent building tools | Kong says it can generate MCP tools and servers from Kong-managed APIs.Source 1 Tools for building agents in Kong AI Gateway are not publicly documented. | Fleet builds agents from a description or a template, without code.Source 11 Code-first agents can be deployed with the LangGraph CLI and app templates.Source 5 |
| Model access | One API to providers including OpenAI, Anthropic, Amazon Bedrock, Gemini, Azure AI, Mistral, and Ollama, with credentials you supply.Source 2, Source 25, Source 57 | Fleet’s managed tiers (Fast, Pro, Max) need no model-provider key of your own; custom models aren’t in that picker.Source 26 LangChain says closed or open models work.Source 3 |
| Integrations and ecosystem | A Policies Hub of policies and integrations.Source 33 AI Vault works with AWS, GCP, Azure, and HashiCorp Vault secrets backends.Source 2 | Fleet has built-in tools such as Gmail, Slack, and GitHub; LangChain says Fleet agents work in Slack, Teams, and Gmail.Source 26, Source 58 |
Which to choose
Choose Kong AI Gateway if
- You want one gateway for LLM, MCP, and A2A traffic, with shared authentication, observability, and policy features.Source 1, Source 2
- You want to govern agents at their existing URLs; Kong says it governs A2A traffic without changing how agents are built.Source 4, Source 49
- You want callers authenticated by API key or your OpenID Connect provider, with per-agent allow or deny lists and rate limits.Source 4, Source 17
- You want one API to model providers such as OpenAI, Anthropic, Gemini, and Amazon Bedrock, switching or combining them without rewriting integrations.Source 25, Source 57
Choose LangSmith if
- You want a platform that builds and runs agents: Fleet builds them without code, and LangSmith Deployment runs agents in production.Source 5, Source 11
- You want to trace agents wherever they run, from any OpenTelemetry-compatible app or through integrations such as OpenAI, Anthropic, and CrewAI.Source 6, Source 29, Source 59
- You need LangSmith’s control plane and Agent Servers in your own infrastructure, for full data residency or air-gapped use, on Enterprise.Source 5
- You want Fleet’s per-agent Clone, Run, and Edit permissions, and human approval before specific agent actions.Source 14, Source 18
Questions buyers ask
Is Kong AI Gateway an alternative to LangSmith?
Partly: both can authenticate calls to agents and handle A2A and MCP.Source 2, Source 4, Source 8, Source 9, Source 10 LangSmith also builds, runs, and traces agents; Kong AI Gateway proxies to agents registered as upstream URLs and routes LLM traffic to major providers through one API.Source 4, Source 5, Source 11, Source 25, Source 28
Do Kong AI Gateway and LangSmith support MCP and A2A?
Kong detects A2A over JSON-RPC and REST bindings and accepts four MCP revisions, the newest from version 2.1.Source 4, Source 48 LangSmith’s Agent Server serves A2A over JSON-RPC only, with documented gaps, and can expose agents as MCP tools; Fleet uses tools from remote MCP servers.Source 9, Source 10, Source 16
Can either one be self-hosted?
Fully self-hosted Kong AI gateways are part of Kong’s separate Gateway Enterprise offering; the 2.x AI entities use a Konnect-managed control plane.Source 2, Source 19 On LangSmith’s Enterprise plan, LangSmith Deployment can run in your own Kubernetes cluster, including air-gapped, and LangSmith comes as BYOC on AWS.Source 5, Source 6, Source 54
How we compare
Read the full methodEvery claim on this page links to a public source. Where none answers a question, the page says so.
We re-check every fact at least every 90 days. This page was last checked .
Something wrong or out of date? Tell us and we’ll correct it.
Sources
66 public sources, each with the date we checked it. Every one opens in a new tab.
Source 2: AI Gateway architecture - Kong Docs Back:abcdefghijklmnopqrst
Source 3: LangSmith: The Agent Engineering Platform Back:abcdefg
Source 4: AI Agents - Kong AI Gateway docs Back:abcdefghijklmnopqrstuvwx
Source 11: No-code agents with LangSmith Fleet Back:abcdefghi
Source 12: How to Build a Multi-LLM AI Agent with Kong AI Gateway and LangGraph Back to text
Source 17: AI Auth Strategies - Kong AI Gateway docs Back:abcde
Source 20: Kong AI Gateway 2.0 Is Now GA - And It's Already Moving Faster Back:abc
Source 21: LangGraph Platform is now Generally Available: Deploy & manage long-running, stateful Agents Back:ab
Source 24: Route A2A agent traffic through AI Gateway - Kong Docs Back:ab
Source 27: Route A2A traffic through AI Gateway - Kong Docs Back:abc
Source 34: Request Termination - Configuration Reference - Policy | Kong Docs Back to text
Source 36: Konnect and Dev Portal audit logs - Kong Docs Back:ab
Source 37: Audit logs Back:ab
Source 38: FIPS 140-3 compliance in AI Gateway - Kong Docs Back:ab
Source 41: Kong AI Gateway changelog - Kong Docs Back to text
Source 44: AI PII Sanitizer - Policy | Kong Docs Back to text
Source 47: Enable additional LangSmith features Back to text
Source 48: MCP version support - Kong AI Gateway docs Back:ab
Source 49: The Agent Gateway for Secure, Observable Agent-to-Agent Communication (Kong) Back:ab
Source 52: Dedicated Cloud Gateways - Kong Docs Back to text