bclaw.sh

Not a script. A teammate.

A claw handles the code reviews piling up. the PRs nobody has time for. the reports due Monday morning. the work you can't get to.

Deploy an AI engineering agent that lives where your team is, runs on your own AWS, and ships real work — code reviews, pull requests, scheduled reports — not just suggestions.

bash
$ npx @boldblackai/create-bclaw swe-pal
/ Hermes / AWS / Slack / Teams / GitHub
Name your claw
@
$ npx @boldblackai/create-bclaw your-claw
Live in your workspace active
SW
swe-pal 8:42 AM
#engineering

Reviewed PR #142 — 3 comments, 1 nit. LGTM once the test is added.

RE
reportclaw 9:00 AM
#reports

Weekly deploy summary: 12 PRs merged, 0 rollbacks. CI avg 4m32s. All green.

SW
swe-pal 3:04 AM
#incidents

Caught a billing spike in us-east-1 at 3 AM. Rolled back the config. Morning numbers attached.

Works where you work

Not a new app to learn.Your existing chat.

bclaw lives inside Slack and Teams. No separate dashboard, no context switch — just @mention and ship.

Slack — BoldBlack
Features

What a claw does, and why it matters

A long-running teammate that lives in Slack and Teams, runs on your own AWS, and ships real work — not just suggestions.

01

It ships real work

Code reviews, pull requests, dashboards, scheduled reports. A claw posts finished work in your channels — not descriptions of what it would do.

02

It lives where your team works

Each claw runs inside Slack, Teams, and GitHub. It reviews code, opens PRs, and responds in the channels your engineers already use every day.

03

It runs while you sleep

Scheduled reports, overnight checks, recurring tasks. A claw wakes up on its own and posts results before your first standup.

04

It remembers everything

State persists on AWS EFS with backups. Your claw keeps its memory, context, and learned preferences across restarts and image updates.

05

It stays under your control

Everything deploys into your own AWS account and runs in one isolated cluster. You own the infrastructure. You decide what it can touch.

06

It takes minutes, not weeks

One command scaffolds the repo. A handful of skills wire up AWS, Slack, and your inference provider. No IT ticket, no setup project, no training videos.

Under the hood, each claw runs hermes-agent on our hardened harness image — deployed to your own AWS account, fully under your control.

Compare

How does a claw stack up?

The honest differences between a claw and the alternatives.

Devin

Devin is a cloud-based autonomous SWE agent — you hand it a task, it works in a sandbox, and you pick up the results.

bclaw

A claw lives in your Slack and Teams channels, runs on your own AWS, and posts finished work where your team already is. No separate dashboard, no sandbox to manage.

Viktor

Viktor is a managed, hosted coding agent with its own infrastructure and pricing tiers.

bclaw

A claw runs entirely in your own AWS account. You own the compute, the secrets, and the data — and you can generate as many claws as you like with one command.

Polfrog

Polfrog is an AI coding assistant focused on code generation within your IDE.

bclaw

A claw is a long-running teammate, not an IDE plugin. It reviews PRs, opens fixes, runs scheduled reports, and works while you sleep — deployed to your own AWS.

A claw does not replace your engineers. It handles the work that piles up — the reviews, the reports, the recurring tasks — so your team ships faster.

Install

From zero to a running claw in five steps

One command scaffolds the repo; a handful of skills take it the rest of the way. Everything runs in your own AWS account.

terminal
$ npx @boldblackai/create-bclaw swe-pal

No name? npm init @boldblackai/bclaw works too — you'll be prompted for one.

  1. 01

    Generate your bclaw repo

    Run the generator with a name for your agent. It scaffolds a complete repository — a set of skills wired up for deployment.

    $ npx @boldblackai/create-bclaw swe-pal
  2. 02

    Wire up AWS + Slack or Teams

    Follow the generated README to create the IAM user and policy for your AWS keys, then create and install the Slack app into your workspace to get its app and bot tokens for .env.

  3. 03

    Open it in your harness

    The generated repo is a set of skills. Open it in Cloud Code, Cursor, or any coding IDE — then let the agent inside. Pi, Hermes, and OpenCode are all supported out of the box.

  4. 04

    Run /setup-bclaw

    The setup skill prompts you for an inference provider. OpenRouter, Z.AI, and Anthropic are supported out of the box, plus anything else hermes-agent supports.

  5. 05

    Manage it over time

    Use /manage-bclaw to update the running image version, edit skills, or tweak SOUL.md. To remove it cleanly, run /teardown-bclaw.

FAQ

Questions, answered

Everything you need to know before you deploy your first claw.

What is bclaw, exactly?
bclaw ("BusinessClaw") is a deployment of hermes-agent configured as a long-running AI agent that lives inside your Slack and Teams channels. It has its own compute on AWS, its own identity, and its own skills. It is not a chatbot — it is an engineering teammate that does real work: reviewing code, opening pull requests, posting scheduled reports, and running tasks you assign.
How is bclaw different from ChatGPT or Copilot?
ChatGPT and Copilot are tools you use interactively — you ask, they suggest. bclaw is an agent that runs autonomously on a schedule. It does not wait for you to type. It wakes up, does the work, and posts the finished result in your Slack and Teams channels. It can open pull requests, review code, and run recurring tasks without human intervention.
What can a claw actually do?
Anything hermes-agent can do, which includes: writing and reviewing code, opening and managing GitHub pull requests, running scheduled reports, monitoring dashboards, answering questions about your repos, and executing multi-step workflows. Each claw is customized through its skills — you define what it knows and what it can touch.
What tools does a claw connect to?
Out of the box, each claw integrates with Slack, Teams, and GitHub. It can also connect to any tool hermes-agent supports, including inference providers (OpenRouter, Z.AI, Anthropic, and more), web search, code execution, and file systems. You extend its capabilities by adding skills.
Where does a claw actually run?
Each claw runs as an ECS task in your own AWS account, inside one isolated cluster. State persists on EFS with backups. Secrets are stored in KMS-encrypted SSM Parameter Store. You own the infrastructure — bclaw just gives you the tools to deploy and manage it.
Is my data secure?
Every claw deploys into your own AWS account with scoped IAM roles, KMS encryption for secrets, and isolated networking. bclaw does not host your data — it lives entirely in your infrastructure. You control access, permissions, and teardown.
How long does setup take?
The scaffold takes one command. Wiring up AWS (IAM user, policy) and Slack (app tokens) typically takes 10–15 minutes following the generated README. After that, you run /setup-bclaw inside your harness to pick an inference provider, and the claw is live.
Can I run multiple claws?
Yes. Each claw is an independent deployment with its own repo, Slack app, AWS stack, and identity. Generate as many as you like — one for code reviews, one for reporting, one for each team. They do not share state unless you configure them to.
What inference providers are supported?
OpenRouter, Z.AI, and Anthropic are supported out of the box. Any provider hermes-agent supports will work — including custom OpenAI-compatible endpoints. You choose during /setup-bclaw.
How do I update or remove a claw?
Use /manage-bclaw to update the running image version, edit skills, or tweak SOUL.md. To remove a claw cleanly, run /teardown-bclaw — it tears down the CloudFormation stack and cleans up resources.
What is hermes-agent?
hermes-agent is the open-source agent runtime that powers each claw. It provides the model orchestration, tool system, skill framework, and platform integrations (Slack, Teams, GitHub, etc.) that make a claw work. bclaw is an opinionated deployment of it.
Do I need to be an AWS expert?
No. The generated README walks you through creating an IAM user, attaching a policy, and copying keys into a .env file. If you can follow a checklist, you can deploy a claw. The infrastructure is production-shaped from the first deploy, so you do not need to design it yourself.
Can multiple people on my team use the same claw?
Yes. A claw lives in Slack and Teams, so anyone in the channels it inhabits can interact with it. You can also deploy separate claws per team or per function — the choice is yours.
What happens if a claw makes a mistake?
A claw operates within the permissions you grant. For code changes, it opens pull requests that go through your normal review process — nothing merges without human approval. You can scope its IAM roles and Slack permissions as tightly as you want.
How does a claw remember things across restarts?
State persists on AWS EFS, which survives container restarts and image updates. The claw's memory, conversation context, and learned preferences are all retained. Backups protect against data loss.
Enterprise

Need to deploy claws acrossyour entire organization?

A single claw is great for one team. When you need dozens — across business units, with centralized governance, audit trails, and sovereign infrastructure — that is Piranesi.

Sovereign

Code, models, and data stay on infrastructure you own. Zero public endpoints.

Compliant

Private networking and full audit trails — designed for HIPAA, SOC 2, and similar.

GitOps

Agents push commits, trigger CI, and deploy through ArgoCD. The full engineering loop.

Piranesi is a BoldBlack platform — the enterprise offering for teams deploying AI agents at scale.

Deploy your first claw today

One command to scaffold it, a few skills to ship it. It runs in your account, on your terms.

bash
$ npx @boldblackai/create-bclaw swe-pal
Every claw ships with
  • Runs on your own AWS account
  • Lives where your team is — Slack and Teams
  • Reviews code and opens PRs
  • Schedules reports and recurring tasks
  • State persisted across restarts
  • KMS-encrypted secrets via SSM