Workforce
The seven named agents are the verbs of the platform. Each agent owns a slice of work, exposes a set of skills, and runs against the tier you pick in the composer. Most days you do not need to think about which agent runs your message. The platform routes for you. When you want a specific shape of work, the slash command picker drops you straight onto the right agent.
Overview
- How many agents
- Seven named, each with a clear domain
- How to invoke
- Slash key in the composer, or plain English
- Skills per agent
- Anywhere from five to two dozen, depending on the agent
- Routing
- Automatic when you do not use a slash command
- Tier applies
- Whatever the picker is set to: Minuet, Allegro, or Forte
- Composability
- Agents call each other when one needs work from another
Think of the seven as colleagues with different jobs. You would not ask the engineer to draft a sales email, and you would not ask the lawyer to research your competitors. Same here. The router picks the right one most of the time. When you want to be explicit, name them.
How to pick an agent
Quick map from the kind of work you want done to the agent that owns it.
| I want to | Use |
|---|---|
| Research a market, a person, a company, a topic | /cortex |
| Write cold emails, follow-ups, LinkedIn DMs | /specter |
| Move a deal forward, qualify, prepare for a call | /striker |
| Draft a blog post, plan a launch, write social copy | /pulse |
| Build a feature, fix a bug, edit code in a repo | /sentinel |
| Publish content across channels, schedule posts | /amplify |
| Draft a contract, review terms, generate an NDA | /counsel |
Cortex
The researcher. Surfaces facts, profiles people and companies, summarises long sources.
Cortex is the agent you call when you need to know something before you act. It runs web searches, reads the result pages, cross-checks the facts, and writes a short brief. Cortex never invents sources. If it cannot find a fact, it says so.
Common use cases
| Ask Cortex to | What you get back |
|---|---|
| Profile a company before a sales call | Snapshot of size, funding, recent moves, key people, and what they care about |
| Research a topic for a blog post | Bullet points of the key arguments with citations for each |
| Find five competitors and compare them | A score card canvas with the dimensions you asked for filled in |
| Read a long PDF and pull the conclusions | A short summary with quoted sections for the parts that matter |
| Investigate a person before a meeting | Background, public posts, anything they have written on the subject |
Specter
The outbound writer. Crafts cold emails, follow-ups, LinkedIn messages, and sequences.
Specter writes outreach in your voice. It pulls the recipient profile from Brain or from a Cortex research run, matches the tone to your brand voice settings, and gives you a draft you can send. Specter also drafts whole sequences: a first touch, two follow-ups, a breakup message, all coherent.
Common use cases
| Ask Specter to | What you get back |
|---|---|
| Write a cold email to a specific person | Subject line plus body, with a hook tied to something specific about them |
| Draft a five-step sequence for a target list | Five messages with timing recommendations and clear variation between them |
| Rewrite this email to be shorter and sharper | A tighter version preserving the asks but cutting filler |
| Follow up on a deal that has gone quiet | Soft, value-led nudge using context from the deal record |
| Personalise this template for fifty leads | A canvas with one row per lead and the personalised opener for each |
Striker
The deal driver. Qualifies prospects, runs discovery, builds proposals, closes.
Striker thinks about deals the way a strong AE does. It looks at the Deals tab, picks up where the last conversation left off, and proposes the next action. It can prepare you for a call, draft a proposal, write follow-up notes that capture what was said, and update the deal record so the pipeline view stays honest.
Common use cases
| Ask Striker to | What you get back |
|---|---|
| Prep me for the call with X tomorrow | A pre-call brief with the recap, open questions, and the ideal outcome |
| Score this deal honestly | Qualification score against the criteria you set, with the weak spots flagged |
| Build a proposal for the project we discussed | A canvas with scope, timeline, pricing options, and the assumptions called out |
| Update the deal record after the call | A clean note summarising decisions, next steps, and the new stage |
| What deals need attention this week | A ranked list with the reason each one needs a touch |
Pulse
The content writer. Long-form posts, social copy, launch plans, newsletter editions.
Pulse writes content meant to be read. Blog posts, LinkedIn threads, newsletter editions, launch announcements. It plans the angle first, drafts the post, and offers variations. Pulse pulls from Brain for facts and voice so every piece sounds like you.
Common use cases
| Ask Pulse to | What you get back |
|---|---|
| Write a 1500-word post on a topic | A full draft with an opening hook, structured sections, and a strong close |
| Plan our next product launch | A launch plan canvas with channels, messaging per channel, and a timeline |
| Turn this blog post into a LinkedIn thread | Six to ten posts, each standalone, with the through-line preserved |
| Draft this week's newsletter | Section-by-section draft pulling from recent activity and a clear theme |
| Give me three headlines for this piece | Three options ranked by likely engagement with the reasoning |
Sentinel
The engineer. Reads code, writes code, edits repositories, runs tests.
Sentinel does the work you would otherwise hand to a senior engineer. It can look at a codebase, propose a change, write the patch, run the tests, and iterate until they pass. It uses the Computer tab to do its work so you can watch it live.
Common use cases
| Ask Sentinel to | What you get back |
|---|---|
| Add a feature to the repo I just cloned | A working patch, the tests that prove it works, and a short summary of choices |
| Fix the failing test on this branch | Root cause diagnosis plus a focused fix that does not touch unrelated code |
| Refactor this module into smaller pieces | A staged refactor with each commit explained |
| Write a script that does X | A self-contained script with a clear command-line interface and a short doc |
| Review my pull request | A read with concrete suggestions, ranked by severity |
Amplify
The publisher. Takes finished content and gets it onto channels.
Amplify is the step after Pulse. Once a post is written, Amplify decides where it should land, formats it per channel, schedules the publication, and reports back. It knows your connected accounts and how each one prefers content shaped.
Common use cases
| Ask Amplify to | What you get back |
|---|---|
| Publish this post to LinkedIn and X | Both posts formatted for each platform with the right length and hashtags |
| Schedule the newsletter for Tuesday 9am | A confirmed scheduled send with a preview link |
| Cross-post my blog post as a Reddit thread | A Reddit-shaped version with the title, body, and chosen subreddit |
| Plan a week of content from this canvas | Day-by-day schedule with the platform per slot |
| Repurpose this video into five short posts | Five posts each highlighting a different beat from the video |
Counsel
The lawyer in the room. Drafts agreements, reviews terms, summarises legal docs.
Counsel is not a substitute for a real lawyer on hard cases. For the every-week stuff (an NDA before a partner conversation, a contractor agreement, terms for a small deal) Counsel writes a clean draft, walks you through the choices, and produces a final document you can send for signature.
Common use cases
| Ask Counsel to | What you get back |
|---|---|
| Draft an NDA between us and a partner | A short, signable NDA with the standard clauses you can review |
| Read this contract and tell me what is unusual | A summary flagging clauses that depart from norm with the risk per clause |
| Generate a SOW from our discovery notes | A statement of work with scope, deliverables, timeline, and payment terms |
| Send this agreement for signature | An agreement record in Brain with the signers, status, and signed PDF when complete |
| What changes between version 1 and version 2 | A side-by-side diff in plain English |
Skills and slash commands
Inside each agent are specific skills you can invoke directly.
A skill is a named unit of work an agent knows how to do. Each agent owns a set. You can browse them by typing the slash key in the composer and then either picking an agent (which exposes all of its skills) or typing the skill name to jump straight to it.
- 01Open the pickerHit forward slash in the composer.
- 02Filter by agent or by skill nameTyping narrows down to matching entries across all agents.
- 03SelectEnter to insert the skill into the composer. The skill prompt template fills in. Add your specifics and submit.
When not to use a slash
Plain English works for almost everything.
You never have to use a slash command. Type your message in plain English and the router resolves to the right agent based on what you asked for. Slash commands matter when you want to be explicit (you know exactly which agent and skill you want) or when you want the skill prompt template filled in for you so you can edit it before sending.