Skip to content

Phyllax Research

Phyllax Research is a cross-integration mode — not a connection to an external service. It runs the AI agentic loop across all your connected integrations at once, letting you ask questions and set up monitors that span tools.

No setup required

There's no credential to enter, no MCP server to spawn, no third-party service. Phyllax Research uses your already-connected integrations together. It just needs 2+ integrations connected.

How it works

When a Phyllax Research schedule runs (or you ask a chat question routed through it), the assistant gets access to every connected integration's MCP tools simultaneously. It can:

  • Read your Gmail and your Stripe at the same time
  • Cross-reference your Slack DMs against your Calendar
  • Compare GitHub PR activity against Linear ticket status
  • Search your Notion docs while querying your HubSpot CRM

The 6 suggested schedules

When you enable Phyllax Research, the wizard offers 6 starting templates:

ScheduleWhat it does
Morning briefingDaily 8am summary of what's new in the last 12h across your tools
Project trackerWatch a specific project name across Slack, Linear, GitHub, Notion
Customer watchCross-integration view of a key customer (Stripe charges + Gmail threads + Slack mentions)
Weekly summaryFriday afternoon recap of the week's key activity
Cross-tool keyword alertWatch for a keyword (your product name, a competitor) across email, Slack, social
End of day recap5pm summary of what you accomplished and what's still open

You can edit any of these or create your own from scratch.

When to use Phyllax Research vs single-integration schedules

Single integrationPhyllax Research
Cost (LLM calls)Cheaper — one MCP server's toolsMore expensive — every connected MCP gets loaded
SpecificityHigh (just Gmail, just Slack, etc.)Lower per-tool, higher cross-tool insight
Best for"Did I get any urgent email?""Has anyone reached out about the Q3 launch?" (could be email, Slack, Linear, etc.)
FrequencyOften every 2–15 minUsually 1–4x/day

Limitations

  • Requires 2+ integrations connected. With one integration, it's the same as a regular single-integration schedule, so the UI doesn't surface it.
  • Token-hungry — each run loads tools from every connected integration, which can be 100+ tools. With smaller LLM context windows, the assistant may not see all of them. Mitigation: keep Phyllax Research schedules at lower frequencies (a few times per day, not every 5 minutes), and use a model with a large context window (Gemini, Claude, GPT-4o).
  • Notification cards for Phyllax Research runs use the search action type — they don't have integration-specific action chips. Click into the card to read the AI's summary.

Troubleshooting

"Less than 2 integrations connected"

Phyllax Research won't enable until you connect at least 2. Go to Integrations, connect a second one (e.g. Gmail + GitHub), then come back.

Result is "Nothing notable"

The assistant scanned everything and didn't find anything urgent — that's the design. Increase frequency or refine the prompt if you want more sensitivity.

Costs spiked

Phyllax Research runs are expensive because they pull tools from every integration. Reduce the frequency, switch the scheduler to Ollama or a cheaper provider in Settings → Advanced, or disable Phyllax Research when you don't need it.

Phyllax — a local desktop sentinel for your work tools