Quiver Docs

Practical product guide for marketers using Quiver day-to-day.

Getting started

Quiver is a GTM command center where AI work is grounded in your product context, not generic advice. Every session starts with your positioning, ICP, competitors, customer research and what worked last time. As you log outcomes, the next session gets smarter.

Create a workspace. Sign up, name your organization and pick a slug. The slug becomes your address: your-slug.quivergtm.dev.
Choose a plan. Founder or Team, 14 days free, card required. Everything is included on both; the only difference is seats.
Connect a model. Add an API key from Anthropic, OpenAI, Google, OpenRouter or any OpenAI-compatible endpoint and pick a default model. Your key, your cost.
Paste your website. One screen: Quiver reads your site and builds your positioning, audience, messaging and competitive landscape in a single pass, then takes you to the dashboard. No website yet? Describe your product in a few sentences instead, or skip and fill the context in by hand.
Start a session. Pick a mode, or describe what you need and let Quiver pick.

Core loop: update context, run sessions, save artifacts and content, log performance, review proposals, repeat.

Workspaces and team

Each workspace is its own subdomain with its own context, team, keys and data. Nothing is shared between workspaces.
Roles: admin manages billing, keys, tokens and team; member uses every feature; anyone else is read-only.
Invite teammates from Settings → Team. Founder includes 3 seats; Team is unlimited.
People can belong to several workspaces and switch from the sidebar; each opens at its own address.

Models and keys

Quiver never proxies or meters your model usage. Calls go straight from Quiver to your provider with your key.
Add connections in Settings → AI providers. Keys are encrypted at rest and never shown again; replace, don't reveal.
Pick a default model, then optionally a different model for each session mode, research processing, performance synthesis and small utility tasks.
OpenRouter is the easy way to mix providers with one key. Any model id works, including ones released after this page was written.
Test connection makes one tiny request and tells you whether the key and model work before you save.

Session modes

Strategy

Use Strategy for positioning work, GTM planning, ICP development, and message architecture. It loads product-marketing-context, marketing-psychology, marketing-ideas, launch-strategy, and competitor-alternatives, then grounds recommendations in your active context and recent learning.

Create

Use Create when you need production-ready marketing assets. You choose artifact type and Quiver loads the matching skill set automatically, while surfacing related past work and performance signals so output quality reflects what already worked for your team.

Feedback

Use Feedback to paste call notes, qualitative reactions, or metrics and get structured synthesis. This mode loads customer-research and is best for turning raw evidence into actions and context update proposals.

Analyze

Use Analyze for performance interpretation and trend reading. It loads analytics-tracking and ab-test-setup, and works well when you need clear interpretation of campaign or channel metrics instead of copywriting output.

Optimize

Use Optimize for CRO, copy critique, and A/B design. It loads page-cro, copy-editing, ab-test-setup, signup-flow-cro, and onboarding-cro to critique existing work against your actual context and ICP, then produce improved versions.

Marketing context

Your context document is the operating system for Quiver. It stores positioning, ICP, messaging pillars, customer language, proof points, active hypotheses, and brand voice.

Use direct edits when you already know what should change.
Use proposal review when updates are AI-suggested from feedback or performance logs.
Every update creates a new version, so you can inspect history and restore older versions.
The active version is injected into every session automatically.
Fields marked Assumption were drafted from your site without direct evidence. Confirm or correct them; saving an edit clears the badge. The dashboard's Context confidence card shows how many are left.

Artifact library

Status workflow: Draft - Review - Approved - Live - Archived.
When an artifact moves to Live, Quiver creates a close-the-loop reminder for later result logging.
Artifacts are versioned; revisions keep historical lineage via parent/child versions.
Artifacts are linked to campaigns for roll-up reporting and context for later sessions.

Campaigns

Campaigns group sessions, artifacts, content, and performance entries into one initiative.
Session and artifact workflows are designed to attach work to campaign scope (`Unassigned` when not specified).
Campaign performance summaries aggregate linked performance logs for quick evaluation.
Use campaign status and priority to separate planning from active execution.

Content

Content in Quiver is a markdown-native store with SEO/OG fields, distribution tracking, repurposing links, and time-series metric snapshots.

Library views: All, Calendar, Drafts.
Repurposing links derived content back to source content.
Metric snapshots track compounding performance over time.

Public API callout

Your published content is available at https://quivergtm.dev/api/public/content/[slug]. This returns markdown body, SEO fields, and OG metadata as JSON. Your website can call it at build time or runtime, so Quiver remains the source of truth.

GET https://quivergtm.dev/api/public/content/[slug]
GET https://quivergtm.dev/api/public/content

Customer research

Research entries support sources such as calls, interviews, surveys, reviews, forums, support tickets, social posts, and Common Room exports.

After save, AI processing runs asynchronously: summary, themes, sentiment, hypothesis signals, extracted quotes, and context proposals.
VoC quote library supports starring quotes; featured quotes are injected into Create and Strategy prompts.
Hypothesis tracking shows whether evidence validates, challenges, or is neutral to active hypotheses.
Push to Linear is a clipboard payload flow (UI and MCP), not a direct Linear API write from Quiver.

Performance log

Log outcomes manually in UI or programmatically through MCP tools.
AI synthesis summarizes what worked and what did not after logging.
Context update proposals can be reviewed and actioned (`pending`, `approved`, `rejected`).
Dashboard close-the-loop queue highlights work awaiting result logging.

Tasks

Every campaign, artifact, content piece and research entry has a task list. The Tasks page shows them all: yours, overdue, due today, grouped by campaign.
Strategy, feedback, analyze and optimize sessions end with proposed next steps. Tick the ones you want, pick an assignee, and they are created and linked to the session.
When an artifact goes live, Quiver creates a "Log results" task due in 14 days for whoever shipped it. That is the close-the-loop queue.
Assignment and due-today notifications are on by default and can be switched off per person in Settings → Notifications.
Tasks are available over MCP too: list_tasks, create_task, update_task, complete_task.

MCP server

The Quiver MCP server lets any MCP client, whatever LLM is behind it, read and write Quiver data directly without using the browser.

Why this matters: your external AI environment can combine Quiver tools with broader context, memory, and other connected systems, while Quiver remains the structured source of record.

Connect Claude.ai (browser)

In Claude, open Settings → Connectors → Add custom connector and paste your workspace URL:

https://<your-workspace>.quivergtm.dev/api/mcp

Use the full address ending in /api/mcp; the workspace root on its own will fail the server check. On the next screen keep what Claude detects: Authentication Sign in now, OAuth client Register automatically, no request headers. Click Add. Claude sends you to sign in to Quiver, you approve the connection, and you are done. The workspace comes from the URL and the user from your sign-in, so you must be a member of that workspace. Everything Claude writes is attributed to you. Each workspace is its own connector: if you belong to two workspaces, add both URLs and approve each once. Admins see every browser connection under Settings → API tokens and can revoke it at any time.

Connect a local client with a token

Clients that can send a static header (Claude Code, Cursor, scripts) can use an API token instead. Create one with the mcp scope in Settings → API tokens; the token decides which workspace the client talks to.

{
  "mcpServers": {
    "quiver": {
      "type": "http",
      "url": "https://<your-workspace>.quivergtm.dev/api/mcp",
      "headers": { "Authorization": "Bearer qvr_..." }
    }
  }
}

Most clients take the same URL and bearer token in their MCP settings. Tokens are scoped to this workspace and can be revoked at any time. Clients that support OAuth can also skip the token and connect the same way as Claude.ai.

Example workflows

"Pull last week's signups and log them to a campaign in Quiver."
"Save this draft as a content piece and track distribution."
"Log this customer call as research and generate a Linear payload."
"Show close-the-loop queue and pending context proposals."
"Propose positioning updates based on this week's evidence."

Full tool list

Context: get_context, get_context_history, propose_context_update, apply_context_update, restore_context_version

Campaigns: list_campaigns, get_campaign, create_campaign, update_campaign, update_campaign_status

Artifacts: list_artifacts, get_artifact, save_artifact, update_artifact, update_artifact_status

Performance: log_performance, get_performance_log, get_close_the_loop_queue, list_proposals, action_proposal

Content: list_content, get_content, save_content, update_content, add_distribution, log_content_metrics, get_content_metrics, get_content_calendar

Research: list_research_entries, get_research_entry, save_research_entry, list_quotes, get_linear_payload

Sessions: list_sessions, get_session

Tasks: list_tasks, create_task, update_task, complete_task

Workspace: get_dashboard_summary

Propose vs apply

Use propose_context_update for suggested changes that need human review. Use apply_context_update only when the human explicitly asks for immediate context mutation.

Billing

Two plans, one product. Founder: up to 3 seats. Team: unlimited seats. Annual billing is two months free.
14-day trial with a card on file. Cancel any time from Settings → Billing; access continues to the end of the period.
Cards, invoices, plan changes and cancellation are handled in the Stripe portal linked from Settings → Billing.
Prefer to run it yourself? Quiver is open source and free to self-host.