Start here

Turn your team’s knowledge into organizational memory.

Connect Codex or Claude Code to your knowledge base, product catalog, and brand voice—shared context your team and agents can use and improve together.

Connect your agent in three steps

1Set up the Senso CLI

Open your terminal and run the commands below. The last one, senso login, asks for an API key — step 2 shows you how to create one.

bash
npm install -g @senso-ai/cli &&
senso skills install quickstart context-layer evaluate-remediate generate-verify publish --global &&
senso login

Continue when you see API key verified.

Installation details

The CLI stores your key for future commands. --global installs the skills for every project rather than only the current one, and with no --agent it installs for every agent found on your machine — Claude Code, Codex, Cursor, Copilot, Gemini and Cline. The five are the four stages of the loop plus the quickstart that sequences them.

2Get an API key Already have a key? Continue to step 3.

3Paste the prompt into your agent

Open Codex or Claude Code and send the prompt below as your first message.

prompt
Use the senso-quickstart skill to get me started with Senso.

The quickstart skill installed in step 1 picks this up and loads the other four as it needs them. Confirm that your agent shows the right organization. Then follow its guidance to add a source or explore your team’s shared context.

Explore what you can do with Senso

What people build with it

Ground an agent

Answers built only from your documents, returned with the sources they came from.

Power a workflow

Support, compliance and call center agents working from one source of truth.

Control what AI says about you

Publish what models should read, then measure whether the answer changed.

Ingest is stage 01 of six

The context layer

One place your agents read from, instead of five places nobody maintains.

Five sections, each holding a different kind of truth about your business. An answer draws on all of them at once, and every sentence it produces traces back to a document you own.

The dots are the readout that matters. A thin section is where answers go vague; an empty one is where they get invented.

How to fill them →Why ground truth →

See it run

Ingest knowledge

Hover any row
Upload your documentsPolicies, rates, termsPDF, Word, Excel, PowerPoint, HTML, CSV, JSON, XML — queryable once compiled.

Happens without you

Read my websitePages read on a scheduleComingYour public pages, re-read as they change. Nobody has to remember to re-upload.
Connect SlackDrop knowledge in a channelComingPost a decision in a channel, it lands in the knowledge base.
Connect your agentsClaude Code · Codex · CLILiveWhatever your agent can read, it can file. One MCP call, no connector to build.

Answer analysis

Running

What products or services do you offer?

10s

Six-week engagements, an annual product catalog1 and the advisory retainer3. Every line is quoted from a document you own2 — add more and run it again.

  1. 1product-catalog_2026.xlsx0.94
  2. 2about-us_who-we-are.pdf0.88
  3. 3Redwood_Six-Week.docx0.71
Sources cited
3
Top relevance
0.94
Answered in

Illustrative until you connect an organization.

Run this on your own knowledgeWith your organization, not an example.Open the playbook →

Already have a RAG system?

Keep it. Ask the same question of your stack and of Senso, then compare. Senso answers only from documents you own, with sources, so the difference is what your system invented or missed. Works against any pipeline, any model, local included.

What this covers, and what it doesn't

MCP covers ingest and query. Generation and publishing are on the CLI. Measurement needs step 2 first.