Anatomy of Glean AI

Jump to: Connectors Enterprise Graph People Assistant Model Hub Skills Agents Sandbox Governance MCP + API

Legend: new Recently launched beta Early access core Core product foundation Infrastructure layer workflow Automation security Governance

📁 glean-workspace/

your AI-powered work operating system

├── 🔌 connectors/

100+ integrations that index your company's data foundation

│ ├── google-drive, slack, jira, confluence

docs, messages, tickets — all indexed and searchable

│ ├── salesforce, github, zendesk, notion

CRM, code, support, wikis — every system of record

│ ├── structured-data (SOQL, JQL, Databricks)

query live structured data with natural language new

│ └── push-api, web-crawler, custom

bring any internal source into Glean's knowledge index

├── 🕸️ enterprise-graph/

the intelligence layer — maps relationships between content, people, and activity foundation

│ ├── entity-extraction

identifies people, projects, products, and topics across all docs

│ ├── relationship-mapping

connects documents to authors, teams, topics, and related content

│ ├── activity-signals

views, edits, shares — learns what's relevant and what's stale

│ ├── personal-graph

per-employee graph of tasks, projects, collaborators, and work style new

│ ├── enterprise-memory

persistent memory layer that retains context across sessions new

│ ├── universal-knowledge

combines enterprise data with real-time world knowledge new

│ └── permissions-layer

every edge in the graph respects source-level access control

├── 👥 people/

understand who works here, what they know, and how they connect org intelligence

│ ├── org-graph

org chart, reporting lines, teams — always up to date

│ ├── expertise-finder

find the right person for any topic based on what they've written and built

│ └── personal-graph

learns your role, projects, and writing style to personalize every answer

├── 💬 assistant/

third-generation AI companion powered by your company's knowledge core

│ ├── chat

natural language Q&A with cited answers from real internal sources

│ ├── search

semantic search across every connected app in one place

│ ├── canvas

collaborative workspace — draft docs, build apps, generate slides

│ ├── real-time-voice

hands-free spoken interaction with the assistant beta

│ ├── branded-slides

AI-generated presentations aligned with your brand guidelines beta

│ ├── proactive-insights

surfaces relevant info before you ask — summaries, action items, reminders new

│ └── permissions-aware-index

users only see what they'd see natively in each source app

├── 🧬 model-hub/

15+ LLMs — choose the right model for every task new

│ ├── llm-choice

switch between OpenAI, Google, Anthropic, Meta, Amazon in real time

│ └── fast-vs-thinking

toggle quick responses vs. extended reasoning per query

├── ✨ skills/

reusable instruction sets — the open Agent Skills standard auto-triggered

│ ├── 📂 personal-skills/

employee-created skills for individual workflows

│ ├── 📂 company-skills/

published, shared skill libraries across the org

│ ├── 📂 ecosystem-skills/

bring in skills from the broader ecosystem new

│ └── SKILL.md

routing metadata + examples — codified once, reused forever

├── 🧠 agents/

automate complex workflows — governed, scoped, with their own identity workflow automation

│ ├── no-code-builder

describe what you want in plain English — visual graph-based editor new

│ ├── agentic-engine-2

next-gen reasoning — adaptive planning, Enterprise Graph-first signals new

│ ├── [agent].instructions

system prompt defining the agent's role, tone, and behavior

│ ├── [agent].data-sources

which connectors & collections this agent has access to

│ ├── [agent].tools

what actions it can take — search, write, call APIs, trigger workflows

│ ├── [agent].triggers

runs on schedule, on event, or when assistant routes to it automatically

│ ├── [agent].identity

name, avatar, suggested prompts — deployed to any team

│ ├── orchestration

event triggers, inter-agent routing, cross-system execution new

│ ├── pre-built-agents

dozens of out-of-the-box agents for common enterprise workflows

│ └── embedded-integrations

run agents in Slack, Teams, Salesforce, Zendesk, ServiceNow, Zoom, GitHub Copilot

├── 🔬 agent-sandbox/

secure, isolated runtime that gives Assistant a virtual computer for multi-step analysis sandboxed

│ ├── AFS (agent-function-service)

sandboxed runtime — isolated pods with filesystem, shell, and persistent state runtime

│ ├── PTC (programmatic-tool-calling)

call tools directly from sandbox code — fewer round trips, faster workflows new

│ ├── code-execution

Python interpreter, shell commands, data transforms — all session-isolated

│ ├── file-system

read, write, upload, download artifacts — persistent across turns

│ ├── pause-resume-snapshots

persist sandbox state across sessions with snapshot & restore

│ ├── web-browsing

retrieves live external context when company knowledge isn't enough

│ └── multi-step-planner

chains search → reason → act into end-to-end autonomous workflows

├── 🛡️ governance/

active data & AI governance — control, compliance, and trust security

│ ├── data-scanning

continuous scanning & remediation of overshared sensitive data

│ ├── infotype-detection

100+ predefined infotypes — SSNs, credentials, PII, custom terms

│ ├── agent-permissions

granular sharing, scoping, and sprawl prevention for all agents

│ └── security-partners

BigID, Cisco, Palo Alto, Rubrik, Virtue AI — open partner program

└── 🔗 mcp + api/

Glean as the system of context — plug in wherever your team already works open platform

├── glean-mcp-server

bring Glean context into Claude, Cursor, ChatGPT, Copilot, and any MCP client

├── mcp-host

Glean agents call external MCP servers for cross-platform actions new

├── mcp-server-directory

20+ curated, validated, ready-to-use MCP servers new

├── native-actions (100+)

read & write to Salesforce, Jira, GitHub, Calendar, and more

├── tool-search

agents auto-identify the most relevant tools based on task intent new

├── rest-api

embed Glean search, chat, and agents into any internal app or workflow

└── 3p-agent-registry