Vernclaw gives AI agents managed access to research, SEO, social, extraction, and generation tools—without building a separate auth, schema, billing, and logging layer for every provider. Use it from a JSON-first CLI, install its agent skills, or connect the same managed runtime through remote MCP. What makes it different: * A stable agent workflow: status → list --json → describe → invoke * Compact JSON output by default; --pretty is for humans
I built Vernclaw after repeatedly hitting the same problem with agent workflows: getting a model to call an API is easy; operating that capability repeatedly is not.
Every useful workflow accumulated more one-off work:
* provider-specific authentication
* changing request schemas
* inconsistent response shapes
* unclear usage and cost
* no shared answer to “which tools are actually enabled?”
* no useful record when a call fails or is denied
Vernclaw is my attempt at a thinner, more operational connector layer.
The CLI is intentionally boring:
```
npm i -g vernclaw-connect-cli
vernclaw-cli login
vernclaw-cli list --json
vernclaw-cli invoke seo.google-trends --keywords "ai agent tools"
```
The default output is compact JSON so an agent or script can parse it reliably. The same active connectors can also be exposed individually through the remote MCP endpoint.
The part I care about most is control, not just connectivity. A connector can be globally disabled, disabled by ID, unavailable to an account until installed or authorized, or blocked when credits are unavailable. Calls still route through the same runtime and leave an activity/audit record.
For ChatGPT, this is a remote MCP connection for Developer Mode—not a claim that Vernclaw is already listed in the public app directory. Claude users can add it as a custom remote connector.
I’d love feedback on two things:
* Which agent workflow are you still rebuilding integration glue for?
* What information would you need before trusting an agent with a managed connector catalog?
Website: https://vernclaw.com/connectors
CLI: https://github.com/hekmon8/vernc...
Report
@mrhkm666 The audit trail piece is what separates this from just another connector wrapper. Knowing a call was denied, and why, matters way more once an agent is acting on your behalf instead of just answering questions.
Report
honestly the workflow design here is really clean. one thing that would be a huge help though is adding a built-in cost tracker per invocation, since you're already handling billing logic. something like a `vernclaw invoke tool --estimate-cost` flag or a running tally per session would save a lot of guesswork when chaining multiple calls together.
Report
Reviews
No reviews yetBe the first to leave a review for Vernclaw
@mrhkm666 The audit trail piece is what separates this from just another connector wrapper. Knowing a call was denied, and why, matters way more once an agent is acting on your behalf instead of just answering questions.
honestly the workflow design here is really clean. one thing that would be a huge help though is adding a built-in cost tracker per invocation, since you're already handling billing logic. something like a `vernclaw invoke tool --estimate-cost` flag or a running tally per session would save a lot of guesswork when chaining multiple calls together.