marketing-team-eve-template: a five-specialist marketing agent team on eve

The problem: marketing work scattered across four tools

Most small teams do not lack marketing tools. They lack a single place where a request turns into finished work. A launch post lives in Typefully, a blog draft in Notion, a campaign in Resend, and the positioning behind all three in someone's head. The README frames the template as the answer to that split: you bring work to a team lead, the lead briefs the right specialist, and the specialist hands back what they produced.

The intended user is a team already deploying on Vercel that wants agent orchestration without writing it. The template ships a lead plus five specialists, each with a defined output. The product marketer owns positioning and the shared brand context document. The content marketer produces long-form and returns a Notion page link. The social media coordinator drafts for X, LinkedIn, Threads, Bluesky and Mastodon into the Typefully queue. The seo specialist audits pages and proposes hierarchy, internal linking and JSON-LD schema. The email specialist adapts copy for the inbox and builds campaigns in Resend.

That division matters because it maps to artifacts, not to chat. Every specialist returns something you can open: a Notion page, a Typefully draft, a Resend campaign link. If your marketing work already lives in those three products, the template removes the copy-paste step between them.

One lead, five specialists, and a single shared document

The architecture is deliberately shallow. The lead loads the brand context and your preferences, writes a brief, and hands back what the specialist produced. According to the README, the lead never writes deliverables itself. That is a real constraint, not a slogan: it means the lead's job is routing and briefing, and quality depends on how well the brief carries context.

Each brief is self-contained. Specialists start fresh with no shared conversation history, so the lead's brief must carry everything, and each specialist reads the brand context document itself. The team shares exactly one piece of state: the brand context document, a short file describing the product, its audience, and what the team claims. The product marketer maintains it and every other specialist reads it at the start of a task. One owner for the only shared state is a clean choice, and it also means a stale brand context document propagates into every specialist's output.

Delegation goes one level deep. Specialists research and edit their own drafts against a written rubric rather than spawning further agents. That caps runaway fan-out and keeps cost predictable, but it also means a specialist cannot hand a sub-problem to another specialist. Newsletters route through two roles: the content marketer writes the prose, then the email specialist adapts it for the inbox and sends it through Resend, so newsletters still get the content marketer's planning and editing passes.

Installing the template and running your first brief

The README points at the one-click Deploy with Vercel button, which provisions the Notion, Resend and Slack connectors, a Vercel Blob store, and prompts for the Typefully API key that the Typefully MCP server uses to read and write social drafts. The connectors set NOTION_CONNECTOR, RESEND_CONNECTOR and SLACK_CONNECTOR, and the Typefully key is stored as TYPEFULLY_API_KEY.

To work locally, link the project you deployed and pull its environment:

vercel link
vercel env pull
pnpm dev          # then run /model once in the TUI to link a provider

That starts the eve dev TUI. The README notes you run /model once inside it to link a provider. The package.json shows the surrounding scripts: pnpm validate runs lint, typecheck and discovery diagnostics in one pass, and npx eve info prints every discovered tool, skill, connection and subagent.

pnpm validate
npx eve info

Before your first email campaign, the README requires two things in the Resend dashboard: verify a sending domain and create at least one segment. The agent sees your verified domains and segments and picks from them, but it cannot create a domain or verify DNS for you. Once that is done, you talk to the team in Slack or the terminal. A fresh Slack conversation pins four starter prompts, one per specialist: sharpen our positioning, write a blog post, draft social posts, review a page's SEO.

Approval gates, and how far they actually reach

The README is specific about what pauses. Sends and deletes in Resend, deletes and scheduled publishes in Typefully, and page moves in Notion all wait for your approval. You get an approve or deny button in Slack or the terminal before the action goes out. Drafting stays friction-free. That split is the right one: the expensive mistake in marketing tooling is an accidental send, not an extra draft.

There is a second boundary worth noticing. The email specialist sees 47 of Resend's roughly 85 tools, so account administration is out of reach entirely. That is a capability restriction enforced by which tools the agent is given, not by a prompt instruction. The full approval matrix, the credential model, and the reasoning behind each boundary are documented in docs/ARCHITECTURE.md, which is where you should look before widening any permission.

The limitation is that approval gates only cover the actions the template knows about. If you connect another tool through MCP, its destructive operations will not inherit this matrix unless you extend the configuration yourself. The README does not document rollback for an approved action that turns out wrong, so treat approval as the last checkpoint rather than a safety net.

Where this template is the wrong tool

The template assumes your marketing output lives in Notion, Typefully and Resend. If your blog runs on a headless CMS, your social scheduling on a different platform, and your email on a provider other than Resend, you are adopting three integrations you will immediately replace. The value here is the wiring, and replacing the wiring removes most of the value.

It also assumes you are on Vercel. The deploy path provisions connectors and a Blob store through the Vercel platform, and local development starts with vercel link and vercel env pull. Running it elsewhere is not described in the README.

Two more boundaries matter. The team shares one brand context document with a single owner, so parallel workstreams that disagree about positioning have nowhere to record the disagreement. And because delegation goes one level deep, a task that genuinely needs a sub-team, such as a full site migration with separate content, SEO and email tracks, will not decompose the way you might expect. You would run it as several separate briefs instead, and the lead has no memory of the earlier ones.

The alternative: wiring the same tools yourself

The obvious alternative is building the same thing directly on eve, or on another agent framework, without the template. The difference is what you inherit. The template gives you a fixed roster of five specialists with defined ownership, a documented approval matrix, a credential model, and a shared brand context document with one owner. Building from scratch gives you freedom over the roster and the integrations, at the cost of designing all of that yourself.

A narrower alternative is a single general-purpose agent with the same MCP connections. That is fewer moving parts and less configuration. What you lose is the routing step: one agent asked to write a blog post and draft social posts tends to blend the two voices, and there is no separate brief forcing a distinct framing per channel. The template's five-specialist split exists to prevent exactly that blending.

A third option is keeping the human workflow and using the tools directly. If your marketing volume is a few posts a month, the setup cost of connectors, a verified sending domain, a segment, and a Typefully API key may exceed what you save. The template earns its place when the routing itself is the bottleneck.

Maintenance, licence and what to check before adopting

The repository is MIT licensed, which permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included. That is the standard permissive position and it is the same licence most Vercel Labs templates carry. It says nothing about the terms of the services the template connects to: Notion, Typefully, Resend, Slack and Vercel each have their own agreements, and the template's licence does not extend to them.

On maintenance, the last push was on 2026-08-20, which is recent. The repository is not archived. There are no releases retrieved, so there is no versioned upgrade path to follow; you track the main branch and re-run pnpm validate after pulling. The dependency set is small and pinned in places: eve at ^0.39.3, ai at ^7.0.70, @vercel/connect at 0.8.0, zod at 4.4.3, and Node 24.x as the engine. The caret ranges on eve and ai mean a minor bump can land new behaviour without a template change, so a lockfile-driven install is the safer default.

Before adopting, check three things. Run npx eve info to see every discovered tool, skill, connection and subagent, and confirm the email specialist really is limited to the 47 Resend tools the README mentions. Read docs/ARCHITECTURE.md for the approval matrix and credential model. And confirm you can verify a sending domain in Resend, because the README is explicit that the agent cannot do it for you.

Editorial conclusion

Adopt it if you already run on Vercel and want marketing output landing in Notion, Typefully and Resend without building the orchestration yourself. Skip it if your stack is outside those tools, or if you need specialists that delegate further down, since the README states delegation goes one level deep. Before trusting it, run npx eve info to see every discovered tool, skill, connection and subagent, and read docs/ARCHITECTURE.md for the approval matrix and credential model.

Frequently asked questions

Official sources

1. Issues
2. License: MIT
3. Project website
4. README
5. vercel-labs/marketing-team-eve-template on GitHub