01
Prepare a commit
Turn the changes already in your repository into a reviewable commit plan.
kb commit generate
A plan to inspect before Git history changes.
Open setupStart with one result
Choose the next useful command.
The launcher asks only for what this outcome needs. You can add the rest later.
02
Prepare a release
Calculate what the next release contains without publishing a package.
kb release plan
A release plan to review before publishing.
Open setup03
Create your own command
Scaffold one focused command for a workflow unique to your team or agent.
kb my-command hello
A discoverable plugin ready for your implementation.
Open setupThe problems
Every team builds the same thing. From scratch.
Your agent went somewhere it shouldn't
Give your agent only the tools you explicitly allow. A plugin without access is simply unavailable.
agentcode-reviewer
✓github.api
✗shell.exec
✓git.read
✗fs.write
✓github.comment
✗network.*
Automation held together with bash
Workflows in YAML. The engine tracks execution, writes logs, and retries on failure.
trigger: pull_request.opened
fetch
✓ 0.3s
review
✓ 2.1s
notify
running…
Switched from OpenAI to Claude — rewrote every LLM call
Plugins call platform.llm.complete(). OpenAI, Claude, Mistral — swap the adapter in config, not a single line in your plugins changes.
my-plugin
platform.llm
.complete(prompt)
.complete(prompt)
adapter
openaiclaudemistral
response
✓ LLMResponse
same interface
Every tool is its own island
Commit, review, release, deploy — one workflow, one platform, one place to look.
code-review.yaml
✓ 3.1srelease.yaml
✓ 12sdeploy.yaml
running…commit.yaml
✓ 0.8sWorkflows
Define once. Run anywhere. Trace everything.
Releases, reviews, QA gates, AI agents — all described as code, running on one engine. Version them in git, run them from CLI or CI, inspect every step.
code-review.yaml
runfetch-diff
✓ 312msgit.diff · pr: {{ env.PR_NUMBER }}
agentreview
✓ 2.8sclaude-sonnet-4-6
✓github.api
✗shell.exec
✓git.read
✗fs.write
condition: score ≥ 7 ?
score < 7
request-changes
github.comment
score ≥ 7
approve
github.approve
agentclaude-sonnet-4-6
$kb github create-pr"fix: rate limiting"
$kb github comment"LGTM, merging"
$kb slack notify"deployed to prod"
plugin commands
plugingithub-plugin
key:ghp_xxxstored in plugin
✓createPR
✗deleteRepo
✓addComment
✗adminAccess
●every call traced · analytics counted
api.github.com
Agents
Agent calls the command. Not your key.
Write a plugin — the agent gets a set of commands. Keys, access rights, audit — inside the plugin. The agent doesn't know how GitHub works — it just calls kb github create-pr.
KB Labs native agents go further — every step is traced, analytics calculated automatically, nothing is lost.
Integrations
21+ adapters. Swap any vendor in config.
Every external dependency — LLM, database, vector store, cache, storage — is behind a typed contract. Change the implementation without touching your code.
LLM
Implement ILLM — any provider works
OpenAI
Databases
MongoDBSQLiteDuckDBRedis
Vector
QdrantVoyage AI
Storage
S3DockerGitHub
Comms
TelegramClickUp
Observability
Pino
Don't see what you need? Write your own adapter — same contract, loads identically.
We run it ourselves
KB Labs runs on KB Labs — every day.
Releases, code review, commit policy, QA gates, dependency management — all running as plugins on the same engine you'll install.
0%
self-hosted
No cloud dependency in the core path
DUAL
license
Full source on GitHub, forever
<1h
first workflow
From install to a working workflow
I build on KB Labs every day. Every release, every review gate, every QA check. It's the same engine you install.
Kirill Baranov · k-baranov.ru· Telegram
FAQ
Frequently asked questions
No. KB Labs is designed to sit beside what you already run. Install it, connect one workflow, and keep everything else untouched. You grow the surface area when it makes sense for your team — never because the tool asked you to.
The OSS core is free and will stay free. It's the engine, and the engine doesn't close. A commercial support layer is on the roadmap for teams that need guarantees — SLA, dedicated support, advanced governance — but it's additive, never a gate in front of the core.
Especially for you. KB Labs is built on the idea that a person plus agents is a team. A plugin-agent can run your reviews. An external agent like Claude or Codex can drive your workflows through the same contracts. You stay in control of what they're allowed to do. Solo builders with agents get the most value from this from day one.
The same way. KB Labs is a runtime with one permissions model, one observability surface, and one extension contract — applied equally to any kind of runner. A human running a workflow, an internal plugin-agent, an external agent, a CI bot — they all declare what they're allowed to do and the runtime enforces it. No one gets privileges. No one escapes the rails. MCP is on the roadmap as the next protocol-level extension of this.
One command installs the launcher. A working environment with your first workflow is typically ready within an hour. No cloud dependency, no provisioning ceremony.