Compare
Squadmates vs building your own Slack AI bot on AWS
A Slack AI bot that a whole company uses needs more than a model call: a routing gate, an identity and owner per bot, access policy, a secrets vault, an audit log, a pause switch, and versioned releases. Squadmates provides these on the Deployments.ai platform, deployed with OpenTofu into an AWS account in your organization.
What it takes
What a production Slack AI bot needs.
A first bot is small: subscribe to mentions, call a model, post a reply. Running bots for a whole company adds the pieces below. Slack and AWS facts link to their own documentation.
| Piece | In Squadmates | Building it yourself |
|---|---|---|
| Routing gate | Every message passes three checks: approved conversation, mentions one teammate, teammate paused. Messages that don't mention a teammate are ignored. Routing gate | Slack sends an app_mention event when someone mentions your bot. You verify each request's signature, decide which channels count, and acknowledge each event within three seconds, or Slack counts the delivery as failed and retries it. |
| Identity and owner per bot | Each teammate gets its own @handle, one job, a named owner, and its own agent deployment. | You design how each job gets its own name, who owns it, and who may change it. |
| Access policy | A permission card per teammate: can read, can remember, can do, must ask. Its name, icon, or instructions never grant access; an admin controls the policy. Permission cards | You write and enforce rules for what each bot reads, stores, and calls. AWS recommends least-privilege permissions in IAM. |
| Secrets vault | Provider keys live in a KMS-encrypted vault, never in a teammate's manifest, and secrets never reach a browser or a model. Models and keys | You store Slack tokens and model keys, for example in AWS Secrets Manager with keys from AWS KMS, and keep them out of code and prompts. |
| Audit log | Each routing decision records who asked which teammate, where, and what happened, without the message text. Audit log | You decide what to log for each request, and how to keep message text out of it. |
| Pause switch | Pause any teammate with one switch. Its credential is revoked, and anyone who mentions it gets a private notice. Pause | You build a way to stop one bot without touching the others, and to revoke its credentials. |
| Versioned releases | Every teammate is a versioned release. Canary a change, roll it back, and conversations in flight keep the version they started on. Releases | You version prompts, tools, and settings, roll changes out, and roll them back. |
On AWS
The same AWS services, already wired.
These are the AWS building blocks Squadmates runs on. If you build your own bot on AWS, each one is a service you set up, secure, and run yourself.
| Building block | In Squadmates | AWS documentation |
|---|---|---|
| Account | An account in your AWS Organization | AWS Organizations |
| API | API on AWS Lambda | AWS Lambda |
| Queues | Amazon SQS queues | Amazon SQS |
| Agent workers | Agent workers on Amazon ECS, in the Deployments.ai runtime | Amazon ECS |
| Data | Postgres and Amazon DynamoDB | Amazon DynamoDB |
| Secrets | AWS Secrets Manager and an AWS KMS vault | AWS Secrets Manager, AWS KMS |
| Models | A model gateway to Amazon Bedrock in your account, or to OpenAI or OpenRouter on your keys | Amazon Bedrock |
| Deploy | Deployed with OpenTofu | OpenTofu |
The honest answer
When building your own is the right call.
Squadmates is not the right answer for every team. Build your own when:
- You need one bot, in one channel, for one job.
- You have no audit log or security review to satisfy.
- You want to own every line, and you have engineers to maintain it.
- You need many integrations now. Squadmates is narrow by design: its Triage teammate drafts replies and has no tools.
- You want to start today without booking an install. Self-hosting Squadmates is coming soon with the source release.
Source
Read the code before you decide.
Squadmates and the Deployments.ai platform are source available under the Functional Source License (FSL-1.1-ALv2), a Fair Source license. The source release is coming soon. Once it is out, you can read the code, self-host it, and change it for your own use, including inside your business. You cannot offer it to others as a competing product or service. Each release becomes Apache 2.0 two years after it is published.
Squadmates is the reference app for Deployments.ai, so everything it does, you can build on the platform. Get notified at release.
Getting it installed
Skip the build, keep the account.
A white-glove install sets up Squadmates and the Deployments.ai platform in an account in your AWS Organization, connects Slack, and recruits your first teammates with you. You own the account and the data. What the install covers.
Comparing more than Slack bots? Read Agentifiers on ways to implement AI agents, or see every way to get AI teammates into Slack.
Book a white-glove install.
Get the routing gate, permission cards, vault, and audit log in your own AWS account, without building them.