CodeGate is live · custom pricing · last verified 24 Oct 2026 · viability monitor
CodeGate
Self-hosted MCP and LLM gateway control plane that ties every AI agent tool call and model call to a real identity on your Kubernetes
Mapping ServiceAccounts and OIDC to MCP permissions is the reason to shortlist this: it answers 'which human caused this tool call' in an audit, which generic API gateways cannot. The Kubernetes assumption is the whole cost of entry, and it does not bend. If your platform team already runs K8s, this is the coherent pick over stitching an API gateway to a separate model proxy.
Verdict generatedVerified 7d ago · liveness 67/100 · cite: rightaichoice.com/tools/codegate
- Platform engineering teams already running fleets of AI agents as Kubernetes workloads
- Enterprises linking LLMs to internal tools that need MCP governance and audit trails
- Security teams that must tie every agent action and tool call to a responsible human
- Companies standardizing AI spend and model selection under one policy plane
- Teams without Kubernetes infrastructure or platform engineering capacity
- Developers who just want a lightweight local proxy for a coding assistant
- Non-technical buyers wanting a GUI-only security tool with no cluster involvement
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Stacklok if you don't run Kubernetes or lack platform engineering capacity to treat agents as workloads your cluster manages.
Self-hosting shifts cost to your own infrastructure — Kubernetes nodes, storage, and the platform engineer time to run the registry, gateways, runtime, and harness add up against the license line.
Stacklok is priced for enterprise platform budgets — procurement runs through sales rather than a self-serve checkout. It sits in the platform-engineering spend category alongside Kubernetes and service-mesh tooling, not in the per-seat productivity-software category. Budget for infrastructure, compute, and the platform engineers who will operate it, not just a license.
In short
CodeGate — Self-hosted MCP and LLM gateway control plane that ties every AI agent tool call and model call to a real identity on your Kubernetes. Best for Platform engineering teams already running fleets of AI agents as Kubernetes workloads, Enterprises linking LLMs to internal tools that need MCP governance and audit trails, Security teams that must tie every agent action and tool call to a responsible human. Pricing is on request (custom quotes).
What is CodeGate?
CodeGate — Self-hosted MCP and LLM gateway control plane that ties every AI agent tool call and model call to a real identity on your Kubernetes. Best for Platform engineering teams already running fleets of AI agents as Kubernetes workloads, Enterprises linking LLMs to internal tools that need MCP governance and audit trails, Security teams that must tie every agent action and tool call to a responsible human. Pricing is on request (custom quotes).
CodeGate is Stacklok's control plane for companies that run AI agents and MCP servers as Kubernetes workloads. It governs two directions of agent traffic: what an agent is allowed to touch (internal services, data stores, SaaS tools) through an MCP Registry and MCP Gateway, and which model it calls and at what cost through an LLM Gateway. The pitch is that an API gateway only proves an agent authenticated, not what it did next, so tool governance and model governance belong in one plane. Identity is the technical core. Kubernetes ServiceAccounts and OIDC claims map to MCP permissions, so each tool call or model call resolves to a person or team rather than a shared token. Two gateways do the securing: an outbound-only tunnel that keeps agent-to-tool traffic behind your firewall, and policy-defined model selection that puts AI spend under your rules instead of the vendor's defaults. The stack comes in layers. ToolHive is the Apache 2.0 open-source MCP platform underneath. Mecatl, announced September 2026, is an open cloud-native agent harness that separates the agent loop from tools and lets teams assemble harnesses from a library of skills, aimed at companies trying to cut agent token costs. Policy travels as code through GitOps, and OpenTelemetry traces span both tool and model calls. Stacklok says the platform now secures 20M+ tool calls per month. Founders Craig McLuckie and Joe Beda co-created Kubernetes, and the company maintains the official MCP registry and is a charter member of the Agentic AI Foundation. The positioning is deliberate: closed stacks from frontier labs and hyperscalers are pleasant until you need to leave. Self-hosted, Kubernetes-native, and identity-first is the trade Stacklok offers. It presumes a platform engineering team that already treats agents as workloads.
Key features
- Run MCP servers as pods inside your own Kubernetes cluster
- MCP Registry for curating and approving trusted internal MCP servers
- MCP Gateway with an outbound-only tunnel for agent-to-tool actions
- LLM Gateway with policy-defined model selection and AI spend control
- Mecatl open cloud-native agent harness for fleets of agents
- Assemble agent harnesses from a library of tools and skills
- Map Kubernetes ServiceAccounts and OIDC claims to MCP permissions
- Unified control plane for agent-to-tool and agent-to-model calls
- Policy-as-code through GitOps for MCP access and agent governance
- Shadow MCP detection to surface unapproved or drifted MCP servers
- Kill switch to shut off compromised or drifted MCP servers
- OpenTelemetry distributed tracing across tool and model calls
- Apply existing Ingress, NetworkPolicy, and service mesh rules to agent traffic
- Self-hosted deployment keeping data and systems behind your firewall
- ToolHive open-source MCP platform core (Apache 2.0)
Behind the verdict
Where Stacklok earns its keep is the seam between tool governance and model governance. Most shops bolt a model proxy onto an API gateway and discover later that the gap in the middle has no owner. Running both gateways from one plane, with the same identity and OpenTelemetry traces, closes that gap in a way that is hard to replicate from parts. Pick this when you have platform engineers who already deploy internal MCP servers as pods and already have an IdP you can trust for OIDC claims. The outbound-only tunnel matters for regulated environments where tool traffic cannot leave the network perimeter. The kill switch argument Stacklok made in October 2026 is the practical one: a server approved in March can start making tool calls nobody recognized in October, and you need a way to shut it off without a cluster teardown. Pass if agents are still a side project. Standing up a registry, two gateways, and a harness is real platform work, and a single developer wanting MCP access for a coding assistant should reach for the open-source ToolHive core instead. We'd also steer non-technical security buyers elsewhere; this is a cluster-native product and reads like one. The closest alternative is the vertically-integrated agent stack from a frontier lab or hyperscaler. Those are genuinely easier to start on. Stacklok's counter is optionality: mix models and providers, swap components, keep data on your own metal. That argument gets stronger as open-weight models close the performance gap, and it gets stronger still when your finance team starts asking what agent spend actually bought. Caveats worth knowing before you commit. Mecatl is new (September 2026), so expect the harness layer to move faster than the gateway layer you are governing production traffic with. The 20M+
Limitations
- Stacklok is built on Kubernetes, so it assumes you already run Kubernetes and have people who can operate workloads there.
- The value comes from policy, identity integration, and observability across a fleet of agents and MCP servers, so it is aimed at enterprise governance rather than individual productivity and pays off mainly at scale.
- Adopters typically assemble and operate multiple Stacklok projects in one stack — ToolHive for MCP servers, Mecatl for the agent harness, plus the MCP Registry, MCP Gateway, and AI Gateway — rather than a single turnkey product.
as of 2026-10-13
Researching CodeGate? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Is CodeGate still active?
CodeGate is live · custom pricing · last verified 24 Oct 2026 · viability monitor.
Viability score
How well maintained and how widely used is CodeGate? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →What's new in CodeGate
Checked 8 days agoAcross the latest 2 updates: 1 launch and 1 news mention.
Why organizations need a kill switch for their MCP servers
Stacklok argues MCP servers approved earlier can later make unrecognized tool calls, so teams need a way to shut off compromised or drifted MCP servers.
Stacklok launches Mecatl open cloud-native agent harness
Stacklok announced Mecatl, an open cloud-native agent harness that separates the agent loop from tools, targeting enterprises reducing agent token costs.
What do users say about CodeGate?
What people actually say about CodeGate — is it worth it?
Our community sentiment research written Jul 3, 2026 is more than 90 days old, so it is being re-reviewed against current data and is not shown.
How we verified this
Every listed fact about CodeGate— pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 24 Oct 2026. Pages mix model-assisted synthesis with human review; see how we verify.
Verification history
We have re-verified CodeGate 11 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — changed pricing page
- — changed pricing page
- — changed who it suits, integrations, summary, our verdict, features, description, who should skip it
- — changed pricing page, limitations
- — changed setup time, switching in, switching out, summary, who should skip it, our verdict, features, pricing fit, hidden costs, description, limitations
- — changed description, summary, our verdict, who should skip it, features, who it suits
Showing the 6 most recent of 11 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does CodeGate cost in 2026?
CodeGate: Pricing is on request (custom quotes). Pricing last verified 24 Oct 2026.
Where the pricing makes sense
The company stage and team size where CodeGate's pricing actually pencils out — and where peers do it cheaper.
Stacklok is priced for enterprise platform budgets — procurement runs through sales rather than a self-serve checkout. It sits in the platform-engineering spend category alongside Kubernetes and service-mesh tooling, not in the per-seat productivity-software category. Budget for infrastructure, compute, and the platform engineers who will operate it, not just a license.
Has CodeGate's pricing changed?
No plan price for CodeGate has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick CodeGate?
Pick CodeGate if you are Platform engineering teams already running fleets of AI agents as Kubernetes workloads, Enterprises linking LLMs to internal tools that need MCP governance and audit trails, Security teams that must tie every agent action and tool call to a responsible human, Companies standardizing AI spend and model selection under one policy plane.
Look elsewhere if you are Teams without Kubernetes infrastructure or platform engineering capacity, Developers who just want a lightweight local proxy for a coding assistant, Non-technical buyers wanting a GUI-only security tool with no cluster involvement, Small teams with no ops bandwidth to run a registry, gateways, and a harness.
Skip it if Skip Stacklok if you don't run Kubernetes or lack platform engineering capacity to treat agents as workloads your cluster manages.
GeneratedReal-world workflow fit
Concrete scenarios for the personas CodeGate actually fits — and what changes day-one when you adopt it.
Install the MCP Runtime and register your internal MCP servers in the MCP Registry, then route agent traffic through the MCP Gateway's outbound-only tunnel so it never leaves the cluster.
Outcome: Agents reach internal databases and APIs with every tool call tied to a ServiceAccount and logged, and no inbound exposure of your network.
Map existing OIDC claims to MCP permissions, turn on OpenTelemetry tracing across tool and model calls, and use shadow MCP detection to surface unapproved or drifted servers.
Outcome: An audit trail that names the responsible human or team for every agent action, plus a kill switch for any MCP server that starts behaving outside approved scope.
Write model-selection and AI-spend policy into the LLM Gateway, and let teams mix frontier and open-weight models through one governed plane.
Outcome: Model choice and spend are consolidated under policy instead of scattered per-team, without locking to a single provider.
Use cases
- Run AI agents with governed access to internal databases and APIs via MCP
- Curate a central registry of approved MCP servers for your organization
- Deploy and manage MCP servers on existing Kubernetes infrastructure
- Enforce policy-as-code on AI tool calls using GitOps workflows
- Monitor MCP tool calls and model calls with OpenTelemetry-based observability
- Tie MCP permissions to existing identity providers via OIDC and ServiceAccounts
- Connect LLMs to enterprise systems while keeping data behind your firewall
- Define model-selection policy to consolidate and govern AI spend
Setup time & first value
How long it actually takes to get something useful out of CodeGate — broken out by persona, not the marketing-page minute.
Plan on weeks, not hours, for a Kubernetes-literate platform team: stand up MCP Runtime, populate the registry, wire the MCP and LLM gateways into your IdP via OIDC, and write the initial GitOps policies. Teams without existing K8s and identity infrastructure should not attempt this as a first project.
Switching to or from CodeGate
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From an ungoverned MCP setup: enroll internal MCP servers in the registry, run them as pods under MCP Runtime, and route agent traffic through the MCP Gateway tunnel.
- →From an API-gateway-only approach: layer the MCP Gateway so agent-to-tool actions are identified rather than only authenticated, and add the LLM Gateway for model-call policy.
- →From frontier-lab agent stacks: replace closed-loop governance with Stacklok's plane and mix models and providers through the LLM Gateway.
- ↗To ToolHive alone: keep the Apache 2.0 MCP platform core and drop the enterprise control-plane layer if you only need MCP server runtime and registry.
- ↗To a frontier-lab or hyperscaler stack: accept the closed loop in exchange for bundled agents, models, and governance.
Integrations
Resources & guides
Tutorials & Learning
YouTube returned 6 videos for “CodeGate”, and we withheld 6: 6 could not be judged, because “CodeGate” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about CodeGate.
Official links
What are the best CodeGate alternatives?
The closest alternatives to CodeGate we list are Codebadger, Context Mode and Context7. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Codebadger | Freemium | Pricing not published | — |
| Context Mode | Freemium | Pricing not published | — |
| Context7 | Freemium | Pricing not published | — |
Alternatives to CodeGate
View all
Codebadger
Codebadger is a self-hosted MCP server that gives AI coding agents graph-based code queries over Joern Code Property Graphs.
Context Mode
Context Mode folds coding-agent tool output before it reaches the model, cutting token spend per request.
Tools that pair well with CodeGate
Common stack mates teams adopt alongside CodeGate, with the specific reason each pairing earns its keep.
Codebadger
Codebadger is a self-hosted MCP server that gives AI coding agents graph-based code queries over Joern Code Property Graphs.
Context Mode
Context Mode folds coding-agent tool output before it reaches the model, cutting token spend per request.
Context7
Context7 delivers version-pinned library documentation into Claude, Codex and Cursor agents so they stop inventing APIs.
CodeGate vs its closest rival
CodeGate's closest head-to-head is with Audioeye — Choose CodeGate if you're a developer or security team wanting to prevent sensitive code leakage when using AI coding assistants. Choose AudioEye if you're an enterprise needing automated web accessibility compliance and legal protection. They solve entirely different problems, so pick based on your primary need: code privacy vs. accessibility compliance. Verdict generated
Head-to-head comparisons
Codegate vs Audioeye
Choose CodeGate if you're a developer or security team wanting to prevent sensitive code leakage when using AI coding assistants. Choose AudioEye if you're an enterprise needing automated web accessibility compliance and legal protection. They solve entirely different problems, so pick based on your primary need: code privacy vs. accessibility compliance.
GeneratedCodegate vs Sublime Security
CodeGate is the clear choice for developers and teams seeking a free, locally-run tool to prevent code leakage when using AI coding assistants. Sublime Security is purpose-built for enterprise security teams fighting email threats like BEC and phishing, but its paid, cloud-only model and lack of code-related features make it irrelevant for CodeGate's domain.
GeneratedCodegate vs Push Security
Choose Push Security if you're a security or identity team needing comprehensive browser protection against AI-driven attacks and shadow AI usage across all browsers. Choose CodeGate if you're a developer or small team solely concerned about code leakage from AI coding assistants and prefer a free, local proxy solution. Push offers richer threat detection but is cloud-based; CodeGate is free and local but limited to code privacy.
GeneratedBito vs Codegate
If you need to govern and secure AI agent access to internal tools on Kubernetes, CodeGate (Stacklok) is the enterprise MCP platform built for that. If your team uses AI coding agents like Cursor or Claude Code and struggles with cross-repo context, Bito’s knowledge graph and AI Architect lift task success rates. Choose CodeGate for infrastructure control; choose Bito for developer productivity at scale.
GeneratedFrequently asked questions
Used CodeGate? Help shape our editorial sentiment research.