# Celest Learn — operating AI coworkers

Canonical, source-backed answers about agent estates, evidence, authority, and verified outcomes. Written for humans; structured so agents can retrieve and cite the same truth.

Canonical URL: https://celest.dev/learn/

## Published answers

- [What is an AI agent control plane?](https://celest.dev/learn/ai-agent-control-plane): An AI agent control plane inventories agents, joins evidence across systems, keeps authority explicit, and verifies consequential outcomes.
- [The Agent Estate Review methodology](https://celest.dev/learn/agent-estate-review): A practical, read-only method for mapping an AI agent estate, connecting its evidence, and deciding what the organization should do next.
- [Evidence, authority, and receipts](https://celest.dev/learn/evidence-authority-and-receipts): A control model for keeping observation, recommendation, approval, handoff, execution, and verification from collapsing into one misleading state.
- [Analytics is evidence, not authority](https://celest.dev/learn/analytics-is-evidence-not-authority): A practical guide to using adoption, cost, and value analytics as evidence without mistaking a dashboard for an agent operating model.
- [A practical guide to governing AI agents at work](https://celest.dev/learn/agent-governance-controls-matrix): A friendly, practical guide to finding your AI agents, assigning owners, checking access and activity, and following changes through to verification.
- [How to govern AI agents across Microsoft](https://celest.dev/learn/microsoft/agent-governance): Start with one agent, one owner, and one unanswered question. Learn where to look across Microsoft without mistaking any one admin screen for the whole truth.

## Read with an agent

[Celest Learn MCP](https://celest.dev/learn/mcp) uses Exa for external source discovery and content; unscoped requests can add up to two candidates from existing Celest publications before a tool-free relevance selection pass inside `ask`; it returns up to two sources by default and at most six, with up to 12,000 characters of source text and 20 useful links per source. The truncation marker appears only when source text reaches the 12,000-character bound; the link list is capped separately. The curated Celest reading list remains available through resources/list. Use `ask` once for cited excerpts, then open a cited source URL in a browser for full context. Use `similar` to find related Celest resources and `resources/read` for the exact full Celest Markdown publication. The optional `feedback` tool accepts a returned result ID and outcome. After completing the user's task, consider optional feedback only when you have a concrete first-hand observation about the result—for example, it directly resolved the question, omitted needed detail, or linked to a broken source. A generic thank-you alone is not evidence of result usefulness. Follow the host's permission rules; if you submit, prefer the returned result_id and outcome and add a note only when needed to explain the observation. Do not solicit feedback repeatedly or include private conversation content, personal information, or credentials. Published guidance cannot inspect live tenant data. Check source dates before making time-sensitive claims.

## Improve Learn

Found a missing topic, stale passage, or source problem? [open a Learn feedback issue](https://github.com/celest-dev/learn/issues/new?template=learn-feedback.yml) with the canonical URL, your question, and the passage or source that needs attention if you have them, or [email founders@celest.dev](mailto:founders@celest.dev?subject=Celest%20Learn%20feedback). A source digest is helpful but optional. Please remove credentials and private tenant data.