Blocks.ai vs Kong AI Gateway
Blocks.ai
Network for AI agents: a free public network, and a private network for each company
Kong AI Gateway
Gateway that governs LLM, MCP, and agent-to-agent traffic
Short answer
Kong AI Gateway is a gateway: it proxies and governs model, MCP, and agent-to-agent traffic, forwarding calls upstream, including to an agent registered as an upstream URL.Source 1, Source 2, Source 3 Blocks.ai is a network agents join by connecting out, with no inbound ports, and on the Pro tier, partners join as their own organizations.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
Blocks.ai
Kong AI Gateway
- Build agents: Not publicly documented
- Host and run agents: Not publicly documented
- Agents across organizations: Not publicly documented
At a glance
What each one is
Blocks.ai
Blocks.ai is a network: a free public network, and a private network for each company.Source 7, Source 15 A company’s private network gives it one place to connect, govern, and audit its agents, whoever built them and wherever they run.Source 15 Blocks.ai does not build, host, or orchestrate agents.Source 11, Source 13
Kong AI Gateway
Kong AI Gateway proxies LLM, MCP, and agent-to-agent traffic through one data plane, with shared authentication, observability, and policies.Source 1, Source 2 In version 2.x, Kong’s Konnect platform manages the control plane and you run data plane nodes.Source 2 Kong announced general availability of 2.0 on 1 September 2026.Source 21
The differences that matter
How agents connect
Blocks.aiEvery Blocks.ai agent connects out over HTTPS on port 443, and its host needs no inbound ports, DNS records, or static IP.Source 5
Kong AI GatewayAn AI Agent entity proxies to an upstream agent URL; in version 2.x, self-managed data plane nodes forward allowed traffic to it.Source 2, Source 3
Kong’s docs example points an agent entity at an internal hostname.Source 3 An outbound-only way for an agent to connect to Kong AI Gateway is not publicly documented.
Model and tool calls
Blocks.aiBlocks.ai doesn’t prescribe an agent’s model, and in its LangChain guide the model client stays in your own process.Source 9, Source 22
Kong AI GatewayKong AI Gateway reaches OpenAI, Anthropic, Amazon Bedrock, Gemini, and other providers through one API, and can turn REST APIs into MCP tools.Source 2, Source 12, Source 23
Governing the model or tool calls an agent makes is not publicly documented for Blocks.ai.
Agents at other companies
Blocks.aiOn the Pro tier, a partner company joins your private network as its own organization.Source 6, Source 7 You can call only the agents it shares with you.Source 16, Source 19
Kong AI GatewayKong AI Gateway proxies to an agent’s URL.Source 3 How another company would control access to its own agents is not publicly documented.
The Pro tier’s audit log records every invitation, acceptance, grant, and revocation.Source 16, Source 24 On Blocks.ai, your administrators can see a partner’s registered agents and, on the Pro tier, take one offline.Source 16, Source 19
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.
| Blocks.ai | Kong AI Gateway | |
|---|---|---|
| What it is | ||
| What it is and who it’s for | A network for AI agents: a free public network, and a private network for each company, with one place to connect, govern, and audit agents wherever they run.Source 7, Source 15 | A gateway that governs LLM, MCP, and A2A traffic, all through one data plane, with shared authentication, observability, and policy features.Source 1, Source 2 |
| Maturity | SDK and CLI 1.0 shipped on 16 June 2026.Source 30 OIDC single sign-on, now on the Pro tier, shipped on 20 July 2026, per the release notes.Source 31, Source 32 | Kong announced AI Gateway in February 2024 and 2.0 general availability in September 2026.Source 21, Source 33 2.2.0 came 30 September 2026.Source 34 Catalog agents are in beta.Source 17 |
| Control | ||
| Agent registry and discovery | On your company’s private network, agents join your private registry.Source 15, Source 35 Pro-tier admins see them all.Source 15, Source 36 Others need a grant to find or call a private agent.Source 19, Source 25 | Each AI Agent entity is scoped to one gateway instance.Source 2, Source 3 An organization-wide agent inventory in Konnect Catalog is in beta.Source 17 |
| Identity and access control | Each agent gets its own identity.Source 14 Only its owner and those granted access by invitation can use a private agent.Source 9, Source 19, Source 25 The Pro tier supports OIDC single sign-on.Source 32 | API key or OpenID Connect through Okta, Azure AD, Google, or any OIDC provider, then a per-agent allow or deny list.Source 3, Source 26 |
| Ownership, policy, and revocation | Pro-tier admins with the right permission can take an agent offline.Source 19, Source 37 Revoked keys and sessions stop working within about 65 seconds.Source 25 | Per-agent, per-consumer, or global policies for input validation, logging, and rate limits.Source 2, Source 3 PII sanitizing covers model requests, as an add-on on Plus.Source 20, Source 38 |
| Audit log and observability | On the Pro tier, an audit log of control-plane changes with who, what, when, and a before-and-after diff.Source 24 Task activity is tracked separately.Source 24 | A2A audit logs record task IDs, method calls, latencies, and errors.Source 8 A2A telemetry can go to Konnect, logging plugins, and OpenTelemetry.Source 3 |
| Connection | ||
| How agents connect | Outbound only, over HTTPS on port 443.Source 5 No inbound ports, DNS records, or static IP on the agent’s host.Source 5 | An AI Agent entity proxies to an upstream agent URL.Source 3 Data plane nodes hold a persistent connection to the Konnect control plane.Source 2 |
| Agents across organizations | On the Pro tier, a partner company joins your private network as its own organization.Source 6, Source 7 You can call only the agents it shares with you.Source 16, Source 19 | Not publicly documented (checked 2 October 2026) |
| Protocol support | An A2A-style task API over JSON-RPC 2.0.Source 39, Source 40 On the free public network, an MCP server lets MCP clients send tasks to agents and manage them.Source 7, Source 41 | A2A over JSON-RPC and REST bindings.Source 3 Four MCP revisions, 2025-03-26 to 2026-07-28 (the last from version 2.1).Source 42 |
| Frameworks, models, and clouds supported | Any agent that takes a task and returns a result, built with any framework and running on your own infrastructure.Source 11, Source 13 SDKs for Node.js and Python.Source 43 | Kong says it governs A2A traffic without changing how agents are built.Source 44 Requests to agents that don’t use A2A pass through as plain HTTP.Source 3 |
| Operations | ||
| Deployment options and data residency | Pro-tier customers each get a single-tenant private instance.Source 15 Agents stay on your infrastructure.Source 11 Enforced data residency is announced, not offered yet.Source 25 | 2.x is hybrid: a Konnect control plane, data plane nodes you run.Source 2 Kong’s pricing also lists Kong-managed Dedicated Cloud and serverless AI gateways.Source 20, Source 45, Source 46 |
| Compliance attestations | In scope under PubNub’s SOC 2 Type II (report under NDA) and ISO/IEC 27001.Source 25, Source 27 Coverage of private instances is not publicly documented. | Kong’s trust center lists ISO 27001:2022 and SOC 2 for Kong Inc.Source 29 From 2.2, a FIPS 140-3 compliant package, not submitted for NIST validation.Source 28 |
| Support and SLA | The Pro tier comes with a 99.999% SLA.Source 7 Security reports are acknowledged within 48 hours.Source 25 Support plans are not publicly documented. | Konnect targets 99.9% a month; Dedicated Cloud Gateways list a 99.99% SLA, serverless AI gateways none.Source 20 Enterprise support SLAs: 30 minutes to 2 hours.Source 20 |
| Time and effort to get running | Ask Blocks.ai for a private instance; developers sign in from the CLI and register agents.Source 15, Source 18 Blocks.ai says the public-network quickstart takes about 10 minutes.Source 47 | A quickstart script creates a Konnect control plane and a local Docker data plane.Source 12 Then you create an AI Agent entity and attach policies.Source 48 |
| Pricing model and public prices | The public network is free.Source 7 Pro pricing is set with each customer.Source 7 | Plus: from $25 a month plus usage; control planes $200 hybrid, $500 dedicated, $25 serverless, per month.Source 20 Enterprise: custom, billed annually.Source 20 |
| Building | ||
| Agent building tools | Blocks.ai doesn’t build or orchestrate agents.Source 11, Source 13 You build with your own framework and write one handler, and the CLI scaffolds, validates, and connects it.Source 9, Source 13 | 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. |
| Model access | Blocks.ai doesn’t prescribe what’s inside an agent, model included.Source 9 In its LangChain guide, the model client stays in your own process.Source 22 | One API to providers including OpenAI, Anthropic, Amazon Bedrock, Gemini, Azure AI, Mistral, and Ollama, with credentials you supply.Source 2, Source 12, Source 23 |
| Integrations and ecosystem | Connection guides for CrewAI, LangChain, LlamaIndex, Microsoft Agent Framework, and n8n.Source 22, Source 49, Source 50, Source 51, Source 52 | A Policies Hub of policies and integrations.Source 53 AI Vault works with AWS, GCP, Azure, and HashiCorp Vault secrets backends.Source 2 |
Which to choose
Choose Blocks.ai if
- Your agents run behind NAT, firewalls, or corporate proxies, where opening inbound ports is hard or not allowed.Source 5, Source 54
- You want partners to join as their own organizations on the Pro tier, and your side to call only what each shares.Source 6, Source 7, Source 16, Source 19
- You want your company’s private registry, where every agent has an owner and starts private, open only to those its owner invites.Source 15, Source 18, Source 19
- You want each agent to have its own identity, with credentials its handler code never sees.Source 14, Source 25
Choose Kong AI Gateway if
- You want one gateway for model, MCP, and agent-to-agent traffic, with shared authentication, observability, and policy features.Source 2
- You need policies on agent traffic itself: input validation, request logging, and per-agent or per-consumer rate limits.Source 3
- You’d rather not change your agents: Kong says it governs A2A traffic without changing how agents are built.Source 44
- You need the control plane in a region you choose, or a fully self-hosted gateway under Kong’s separate Gateway Enterprise offering.Source 20, Source 55
Why teams choose Blocks.ai
One registry, with an owner for every agent
On your company’s private network, each agent connects out with no inbound ports and joins your private registry, wherever it runs.Source 5, Source 15, Source 35 Kong agent entities each belong to one gateway; Konnect Catalog’s inventory is in beta.Source 2, Source 17
No inbound ports where agents run
Blocks.ai agents stay where they run, on a cloud VM or corporate network, with no inbound ports.Source 5, Source 11 Kong AI Gateway proxies to each agent’s URL.Source 3 Kong doesn’t publicly document an outbound-only path for agents.
Questions buyers ask
Do Kong AI Gateway and Blocks.ai support A2A and MCP?
Kong AI Gateway detects A2A requests over JSON-RPC and REST and accepts four MCP revisions, 2025-03-26 to 2026-07-28 (the last from version 2.1).Source 3, Source 42 Blocks.ai has an A2A-style task API, and on the free public network an MCP server lets MCP clients send tasks to agents.Source 7, Source 39, Source 41
Do agents need changes to work with Kong AI Gateway or Blocks.ai?
Can Kong AI Gateway be self-hosted?
Yes, as a separate offering: fully self-hosted AI gateways are part of Kong’s Gateway Enterprise.Source 20 The 2.x AI entities are hybrid only: Konnect manages their control plane, and you run the data plane nodes.Source 2 Self-hosted Kong Gateway configures the same capabilities with AI plugins.Source 56
How are Kong AI Gateway and Blocks.ai priced?
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
56 public sources, each with the date we checked it. Every one opens in a new tab.
Source 2: AI Gateway architecture - Kong Docs Back:abcdefghijklmnopqrstuvw
Source 3: AI Agents - Kong AI Gateway docs Back:abcdefghijklmnopqrstuv
Source 8: Route A2A traffic through AI Gateway - Kong Docs Back:abc
Source 15: Your company's private network Back:abcdefghijklmn
Source 21: Kong AI Gateway 2.0 Is Now GA - And It's Already Moving Faster Back:ab
Source 26: AI Auth Strategies - Kong AI Gateway docs Back:ab
Source 27: Security Back:ab
Source 28: FIPS 140-3 compliance in AI Gateway - Kong Docs Back:ab
Source 33: Announcing Kong’s New Open Source AI Gateway with Multi-LLM Support, No-Code AI Plugins, Advanced Prompt Engineering, and More Back to text
Source 34: Kong AI Gateway changelog - Kong Docs Back to text
Source 38: AI PII Sanitizer - Policy | Kong Docs Back to text
Source 42: MCP version support - Kong AI Gateway docs Back:ab
Source 44: The Agent Gateway for Secure, Observable Agent-to-Agent Communication (Kong) Back:abc
Source 45: Dedicated Cloud Gateways - Kong Docs Back to text
Source 48: Route A2A agent traffic through AI Gateway - Kong Docs Back to text
Source 51: Connect Microsoft Agent Framework to Blocks Back to text
Source 53: Kong AI Gateway Policies Hub - Kong Docs Back to text
Source 54: Blocks Network Architecture whitepaper Back to text
Source 56: Configure Kong AI Gateway on-prem - Kong Docs Back to text