GitHub delivery evidence
Build and verify the change with GitHub context alongside it.
Import your OpenAPI specs, watch the surface change revision by revision, and run real requests with per-environment secrets through a proxy that refuses to hit anything internal.
openapi.yaml · endpoints
One connected loop, held on the stage this capability serves. The other stages stay as context so you can see what feeds in and what comes next.
Build and verify the change.
Build and verify the change with GitHub context alongside it.
Inspect delivery workflow activity in the connected release workspace.
Investigate test outcomes and flakiness evidence across runs.
Keep API contracts inspectable next to the release that changed them.
Compare public cloud compute pricing as planning evidence.
Built for OpenAPI
This is a REST workbench. OpenAPI in, real requests out, one call at a time. Git-sourced specs come from GitHub. No GraphQL, no gRPC, no multi-step runners. We do one thing here and we do it properly.
Every sync writes a numbered, immutable revision. Compare any two snapshots and see exactly how the contract moved: endpoints added, operations newly secured, deprecations.
commit 8f2c1a4 · content changed
sha256 3ab9…e71
commit 41d0b9c · content changed
sha256 77c2…10a
commit a19e6f2
The server-side proxy is permission-gated, resolves and checks the target against SSRF and DNS rebinding, bounds the body and redirects, and writes every execution to the audit log.
every execution is permission-gated and written to the audit log
Import once, then it keeps up
Bring a spec in three ways, and the workbench does the rest: it validates the document, snapshots the stats, tracks each revision, and — for GitHub-sourced specs — auto-syncs on every push. From the same place you read an operation, fill its variables and run it.
Point at a hosted document, push a file, or connect a GitHub repo where a webhook auto-syncs every push. Each source is validated before it lands.
Fetch from a URL
Point at a hosted openapi.json or .yaml; re-fetch on demand or on a schedule.
Upload a document
Push raw JSON or YAML straight in — from the dashboard or the uploader CLI.
Connect a repo
Track a spec at a path on a branch; a webhook auto-syncs every push.
Named environments scope to org, team, or a single spec, with typed string, url, and secret variables. Secrets are encrypted at rest and never rendered back.
secret · encrypted at rest, never rendered
Each revision records the numbers that matter — endpoints, secured operations, deprecations, models, request bodies — and the operations behind every tag.
47
endpoints
18
secured
3
deprecated
29
models
22
request bodies
9
tags
operations by tag
Open an endpoint to read its params, body, responses, and security; fill the variables and run it; then check the recent runs, all in one pane.
Overview
params, body, responses, security
Request
fill variables, then execute
Recent
your last runs for this operation
The published uploader sends a fresh document on every build, so the tracked contract never drifts from the code. It runs in GitHub Actions or any generic CI.
@lubed/api-spec-uploaderci · pipeline
npm i -g @lubed/api-spec-uploader
lube-openapi --api-spec-id spec_9f2 --file openapi.yaml --message "release v3.2"
rev 16 · endpoints 47 · content changed
Someone in customer success needs to know if there’s an endpoint for bulk refunds. Today that means pinging an engineer, who then digs up the docs, or worse, turns on a tool and walks them through it. The answer exists. Getting to it costs two people an afternoon.
MCP Explorer will expose your tracked specs as a Model Context Protocol server, so a teammate’s AI assistant can search the surface, read the schema, and draft the call. Share a spec with a link and they explore it directly. Every request still runs through the same permissioned, SSRF-hardened proxy, so self-serve never means unsafe. It’s on the roadmap — the tracked specs and the safe proxy it builds on are here today.
Old way · waits on a dev
“Which call handles refunds? Can you enable the docs for me?”
With lube · self-serve
Will ask the MCP Explorer, find the endpoint, and run it in the sandbox.
Search the surface
Ask in plain language; the MCP server returns matching endpoints and schemas.
Share a spec link
Hand a colleague a link instead of a screenshot of the docs.
Run it safely
Calls still pass permissions and the SSRF-safe proxy, and land in the audit log.
Import sources, sync triggers, variable types, and per-revision stats are all first-class, so the contract keeps up with the code.
Spec sources
Three ways in.
Sync triggers
How a re-check fires.
Variable types
With scopes.
Methods tracked
Counted per revision.
Spec stats
Health over time.
Proxy limits
Safety by default.
Bring one OpenAPI document and start.