Build me a Slack agent on Cantelop that monitors production. Cantelop is a hosted runtime where each agent session runs in its own microVM with a workspace that outlives it.
Use the Cantelop SDK: https://github.com/stepandel/cantelop-sdk
Docs for reference: /docs
Before you write any code, ask me:
- Which channel does it live in, and should it respond only when mentioned or follow threads too?
- Where do alerts come from — Alertmanager, PagerDuty, Grafana, something else — and what does a payload look like?
- What can it read about production: health URLs, a logs or metrics endpoint, deploy history?
- Should it only report, or may it act — restart, roll back, page someone?
- Whose Workspace is it (if there's more than one user or environment to the agent): a user, a customer account, a team, a codebase? Cantelop creates workspaces that could be shared between sessions, so pick the identity that owns the files.
- Which agent SDK should the runtime use: openai, claude, pi, or something else?
Then scaffold with `cantelop init -app ops-bot -provider <my answer>`, keep the API (src/api.ts) and the Session runtime (src/session.ts) separate as the template does, run it with `cantelop dev`, and tell me which env vars and secrets to set and where to point the webhook after `cantelop deploy`.
Build me an email agent on Cantelop that handles customer support. Cantelop is a hosted runtime where each agent session runs in its own microVM with a workspace that outlives it.
Use the Cantelop SDK: https://github.com/stepandel/cantelop-sdk
Docs for reference: /docs
Before you write any code, ask me:
- Which inbound email provider — Postmark, SendGrid, Resend, something else — and which address?
- Where does product knowledge come from: a docs folder, a help center, past tickets?
- When must it hand off to a human, and how: a draft for review, a forward, a ticket?
- What tone and signature, and should it send directly or queue replies for approval at first?
- Whose Workspace is it (if there's more than one user or environment to the agent): a user, a customer account, a team, a codebase? Cantelop creates workspaces that could be shared between sessions, so pick the identity that owns the files.
- Which agent SDK should the runtime use: openai, claude, pi, or something else?
Then scaffold with `cantelop init -app support-agent -provider <my answer>`, keep the API (src/api.ts) and the Session runtime (src/session.ts) separate as the template does, run it with `cantelop dev`, and tell me which env vars and secrets to set and where to point the webhook after `cantelop deploy`.
Build me a chat agent on Cantelop that's embedded in my site. Cantelop is a hosted runtime where each agent session runs in its own microVM with a workspace that outlives it.
Use the Cantelop SDK: https://github.com/stepandel/cantelop-sdk
Docs for reference: /docs
Before you write any code, ask me:
- Which site and pages, and how should it be embedded: a script tag, an iframe, a React component?
- What should it answer from: a sitemap, a list of URLs, a docs repo?
- Should it remember a visitor across visits, and for how long?
- What should it refuse or hand off — pricing, account questions, anything the docs don't cover?
- Whose Workspace is it (if there's more than one user or environment to the agent): a user, a customer account, a team, a codebase? Cantelop creates workspaces that could be shared between sessions, so pick the identity that owns the files.
- Which agent SDK should the runtime use: openai, claude, pi, or something else?
Then scaffold with `cantelop init -app site-chat -provider <my answer>`, keep the API (src/api.ts) and the Session runtime (src/session.ts) separate as the template does, run it with `cantelop dev`, and tell me which env vars and secrets to set and where to point the webhook after `cantelop deploy`.
Build me a QA agent on Cantelop that I can tag in a GitHub PR. Cantelop is a hosted runtime where each agent session runs in its own microVM with a workspace that outlives it.
Use the Cantelop SDK: https://github.com/stepandel/cantelop-sdk
Docs for reference: /docs
Before you write any code, ask me:
- Which repositories, and how is it summoned: a mention in a PR comment, every new PR, a label?
- What should it run: the project's tests and lint, a build, a manual walkthrough of the change?
- How should it report: a review with inline comments, one summary comment, a check run?
- May it approve or request changes, or only comment?
- Whose Workspace is it (if there's more than one user or environment to the agent): a user, a customer account, a team, a codebase? Cantelop creates workspaces that could be shared between sessions, so pick the identity that owns the files.
- Which agent SDK should the runtime use: openai, claude, pi, or something else?
Then scaffold with `cantelop init -app pr-qa -provider <my answer>`, keep the API (src/api.ts) and the Session runtime (src/session.ts) separate as the template does, run it with `cantelop dev`, and tell me which env vars and secrets to set and where to point the webhook after `cantelop deploy`.
Build me a Linear agent on Cantelop that triages tickets. Cantelop is a hosted runtime where each agent session runs in its own microVM with a workspace that outlives it.
Use the Cantelop SDK: https://github.com/stepandel/cantelop-sdk
Docs for reference: /docs
Before you write any code, ask me:
- Which teams and projects, and which events: new issues only, or edits too?
- What may it set: priority, labels, project, assignee, estimate?
- What are the team's conventions for those, or should it learn them from past issues?
- Should it comment with its reasoning, and how do humans correct it when it's wrong?
- Whose Workspace is it (if there's more than one user or environment to the agent): a user, a customer account, a team, a codebase? Cantelop creates workspaces that could be shared between sessions, so pick the identity that owns the files.
- Which agent SDK should the runtime use: openai, claude, pi, or something else?
Then scaffold with `cantelop init -app linear-triage -provider <my answer>`, keep the API (src/api.ts) and the Session runtime (src/session.ts) separate as the template does, run it with `cantelop dev`, and tell me which env vars and secrets to set and where to point the webhook after `cantelop deploy`.
Build me a Telegram agent on Cantelop with access to my documents. Cantelop is a hosted runtime where each agent session runs in its own microVM with a workspace that outlives it.
Use the Cantelop SDK: https://github.com/stepandel/cantelop-sdk
Docs for reference: /docs
Before you write any code, ask me:
- Where do the documents live: a git repo, a bucket, a folder of URLs, files sent to the bot?
- Who may use it: one private chat, a group, anyone with the link?
- Should answers quote their source, and should it say when the documents don't cover a question?
- Should documents sent to the bot be added to the set?
- Whose Workspace is it (if there's more than one user or environment to the agent): a user, a customer account, a team, a codebase? Cantelop creates workspaces that could be shared between sessions, so pick the identity that owns the files.
- Which agent SDK should the runtime use: openai, claude, pi, or something else?
Then scaffold with `cantelop init -app docs-bot -provider <my answer>`, keep the API (src/api.ts) and the Session runtime (src/session.ts) separate as the template does, run it with `cantelop dev`, and tell me which env vars and secrets to set and where to point the webhook after `cantelop deploy`.
Build me an agent on Cantelop for my own use case. Cantelop is a hosted runtime where each agent session runs in its own microVM with a workspace that outlives it.
Use the Cantelop SDK: https://github.com/stepandel/cantelop-sdk
Docs for reference: /docs
Before you write any code, ask me:
- What job does it do, and what does a good outcome look like?
- What wakes it: a webhook from which service, a chat message, a call from my own app, a schedule?
- What does one conversation belong to: a user, a thread, a ticket, a repository?
- What does it need access to, and what should it remember between requests?
- What must it never do without a human?
- Whose Workspace is it (if there's more than one user or environment to the agent): a user, a customer account, a team, a codebase? Cantelop creates workspaces that could be shared between sessions, so pick the identity that owns the files.
- Which agent SDK should the runtime use: openai, claude, pi, or something else?
Then scaffold with `cantelop init -app my-agent -provider <my answer>`, keep the API (src/api.ts) and the Session runtime (src/session.ts) separate as the template does, run it with `cantelop dev`, and tell me which env vars and secrets to set and where to point the webhook after `cantelop deploy`.