For research and analyst teams

Every analyst's AI follows your method and cites your sources.

Research quality depends on method and sources. When each analyst's AI improvises both, reports stop being comparable. Put the approved method, source list and templates where every agent reads them.

Get PromptOwl free Get a demo

In their words

What research and analysis teams say when AI starts to drift.

Each of these has the same cause: the knowledge your AI needs has no governed home that every agent reads from. Find your problem →

“Two analysts' AIs cited different sources for the same number.”
“The AI used a source we stopped trusting.”
“Every report is formatted differently.”
“We can't rebuild how last year's report was produced.”

Your team's shared brain

What goes in, and the agents that run on it.

What goes in

Research methods and frameworks
A catalog of trusted sources, and the ones you've retired
Report templates
Past findings, versioned

Each area gets an owner who approves changes, so agents read the version your team signed off on.

Research assistant

Gathers material from approved sources only.

Never sees: Sources marked as retired.

Report drafter

Drafts in the approved template and method.

Never sees: Unpublished client findings from other engagements.

Get started

Three steps to a working start.

1
Start a nest from the analyst starter
Folders for methodologies, sources and reporting, with a research framework and report template.
Terminal
```
$ npx @promptowl/contextnest-cli init --starter analyst
```
Get PromptOwl free →
2
Approve your source catalog
List the sources you trust and retire the ones you don't. Agents stop citing a retired source.
3
Connect your agent and ask it to map how it can help
In Claude Code, add the ContextNest plugin (or use the Connect button in a hosted nest), then paste the prompt below. It reads your nest, proves it, and drafts an opportunity map for you to approve.
In Claude Code
```
$ /plugin marketplace add promptowl/contextnest
```
Then
```
$ /plugin install contextnest
```

Map how it can help

Ask your agent what it could do for your team.

Paste this into Claude Code once it's connected to your nest. The agent proves it can read your knowledge, then drafts an opportunity map for you to approve.

Claude Code · a fictional support team's nest

Example

1. Prove it
2. Get oriented
3. Map it
4. Save as a draft

The prompt

My team is the research and analysis team.

You're connected to my ContextNest nest. Before anything else, prove it: search the nest, open two or three documents, and tell me their titles and one line from each. If you can't read real documents, say so and stop. Don't guess what's in the nest.

Use search and single-document reads. If one way of reaching the nest fails, try the other tools you have (the ContextNest MCP tools, the ctx command line or the REST API) before you give up.

If the nest is empty or nearly empty, ask me three questions first: what my team does, which questions people ask us most, and which documents we'd trust an agent to answer from. Then work from my answers.

Then build me an opportunity map:
1. Five things you can do for me today with what's in the nest. For each one, name the documents it relies on.
2. What's missing: the documents or decisions that would let you do three more things, and who on my team probably owns each one.
3. The agents you'd run on this nest: what each one does, and what each one should not be allowed to see.

Save the map back to the nest as a draft titled "How AI can help us: opportunity map". Don't publish it or share it anywhere. Ask me before you publish anything.

One shared brain

Every team adds its own area.

Start with one team. The next team starts from what's already approved.

See every team See how it can help

Start with your team. Free.

Connect the AI tools you already use, and approve what they're allowed to rely on.