Enterprise Agents

Drop-in OpenClaw / NemoClaw agent configuration for the ibl.ai Enterprise solution segment — one Workplace Assistant orchestrator plus 12 specialist subagents, packaged so a NemoClaw host can import the whole roster with zero conversion.

---

This directory is a complete, drop-in OpenClaw/NemoClaw configuration for the Enterprise solution segment. It contains one parent orchestrator agent and 12 specialist subagents covering every major enterprise workplace function.

Copy the contents of this directory directly into /sandbox/.openclaw/ on a NemoClaw sandbox host and the configuration is immediately operational.

---

Agent Roster

Agent ID	Name	Role
enterprise-assistant (default)	Workplace Assistant	Segment entry point — greets employees, interprets intent, and delegates to specialist subagents
knowledge-agent	Knowledge Search	Enterprise search across internal wikis, documentation, and knowledge repositories
it-help-desk-agent	IT Support	Issue resolution, password resets, software access provisioning, and ticket management
sales-enablement-agent	Sales Navigator	Competitive briefs, deal strategy, prospect research, and outreach templates
customer-support-agent	Customer Support	Ticket resolution, customer account management, escalations, and follow-up communications
onboarding-agent	Onboarding Guide	New hire orientation, ramp-up planning, system access setup, and first-week navigation
hr-agent	HR Assistant	Policy Q&A, benefits guidance, leave management, and people operations support
marketing-agent	Marketing Strategist	Content creation, campaign planning, SEO guidance, and brand-aligned collateral
engineering-agent	Engineering Advisor	Code review, technical documentation, architecture guidance, and project onboarding
meeting-agent	Meeting Scribe	Meeting recaps, action item extraction, follow-up drafts, and calendar coordination
training-agent	Learning Coach	Employee development paths, compliance training deadlines, and course recommendations
data-analysis-agent	Data Analyst	Business reports, trend analysis, dashboard interpretation, and ad-hoc queries
operations-agent	Ops Optimizer	Workflow automation, process bottleneck identification, vendor management, and SLA tracking

---

Directory Layout

enterprise/
├── agents/                        # One directory per agent (13 total)
│   ├── enterprise-assistant/      # Default entry-point orchestrator
│   └── <subagent-id>/             # 12 specialist subagents
│       └── agent/
│           ├── IDENTITY.md        # Required: agent persona (name, role, vibe)
│           ├── SOUL.md            # Required: behavioral guidelines
│           ├── TOOLS.md           # Required: available integrations, usage notes, and data sources
│           ├── auth-profiles.json # Required: Anthropic API key for this agent
│           ├── AGENTS.md          # Optional: subagent routing table (parent only)
│           ├── USER.md            # Optional: user/environment-specific context (rare)
│           ├── BOOTSTRAP.md       # Optional: one-time first-run setup actions
│           ├── HEARTBEAT.md       # Optional: periodic monitoring checklist (schedule-driven agents)
│           └── MEMORY.md          # Optional: seed long-term facts (policy/regulation-heavy agents)
├── skills/                        # Third-party platform skill definitions
│   └── <tool>/
│       └── SKILL.md               # Skill definition with frontmatter and body
├── workspace/                     # Shared workspace files
├── openclaw.json                  # Segment-level gateway configuration
├── .env.example                   # Sample OpenClaw daemon env file (~/.openclaw/.env)
└── README.md

---

Architecture

enterprise-assistant (default, entry point)
├── knowledge-agent
├── it-help-desk-agent
├── sales-enablement-agent
├── customer-support-agent
├── onboarding-agent
├── hr-agent
├── marketing-agent
├── engineering-agent
├── meeting-agent
├── training-agent
├── data-analysis-agent
└── operations-agent

The parent agent (enterprise-assistant) receives all incoming employee messages. It interprets intent using its AGENTS.md routing table and delegates to the appropriate specialist via sessions_spawn. For requests that span multiple domains (e.g., "set up access for my new hire and send them an onboarding checklist"), it spawns multiple subagents in parallel and synthesizes the results into a single response.

---

Tool Skills

Each skill in skills/ is a directory containing a SKILL.md file that describes a third-party platform one or more agents in this segment can integrate with — what the platform is, how agents use it, the API operations available, credential requirements, and scope or permission notes. Agents reference the relevant skill documents in their TOOLS.md to understand how to interact with each system.

All skills draw their runtime credentials from the OpenClaw daemon env file at ~/.openclaw/.env on the sandbox host. The .env.example file checked into this repository is the committed sample template; every value in it is a non-functional placeholder. Copy it to ~/.openclaw/.env and fill in real API keys before deploying (see import instructions below).

Skill directory	Platform description
confluence/	Confluence team wiki and documentation platform - lets an agent search, read, and create pages across internal knowledge bases, runbooks, technical documentation, and policy spaces.
github/	GitHub source control and collaboration platform - lets an agent read pull requests, commits, repository metadata, and CI workflow results to support engineering review and documentation.
hubspot/	HubSpot CRM and Marketing Hub - lets an agent read and update deals, contacts, companies, marketing campaigns, and workflow enrollments across the inbound sales and marketing lifecycle.
jira/	Jira project and issue tracking - lets an agent create, query, update, and link issues, sprints, and epics across engineering, IT, and operations projects.
okta/	Okta workforce identity platform - lets an agent verify employee identity, look up user profiles and group memberships, check MFA enrollment status, and support access provisioning workflows.
salesforce/	Salesforce CRM and Sales Cloud - lets an agent read and write opportunities, accounts, contacts, cases, and activities across the enterprise revenue and support lifecycle.
servicenow/	ServiceNow ITSM and enterprise service management - lets an agent create, read, update, and route incidents, service requests, HR cases, change records, and knowledge articles.
slack/	Slack team messaging platform - lets an agent post messages, send direct messages, read channel history, and deliver notifications into employee workflows.
snowflake/	Snowflake cloud data warehouse - lets an agent execute read-only SQL queries, explore table and schema metadata, and retrieve business intelligence results for reporting and analysis.
workday/	Workday HCM - lets an agent retrieve employee records, leave balances, onboarding task status, job history, and compensation data in read-only mode to support HR and onboarding workflows.
zendesk/	Zendesk customer service platform - lets an agent create, read, update, and search support tickets, manage user records, apply macros, and retrieve CSAT survey results.
zoom/	Zoom video conferencing platform - lets an agent retrieve meeting metadata, participant lists, cloud recording URLs, and auto-generated transcripts to support meeting intelligence workflows.

---

NemoClaw Import Instructions

Follow these steps to deploy this configuration on a NemoClaw sandbox host.

1. Copy the directory contents into the sandbox root.

git clone https://github.com/iblai/enterprise-agents.git
cp -r enterprise-agents/. /sandbox/.openclaw/

All files must land at /sandbox/.openclaw/ — not inside a subdirectory of it.

2. Set correct ownership on agent directories.

chown -R sandbox:sandbox /sandbox/.openclaw/agents/
chown -R sandbox:sandbox /sandbox/.openclaw/workspace/

3. Replace the sample credentials in every auth-profiles.json.

Each agent directory contains /sandbox/.openclaw/agents/<agent-id>/agent/auth-profiles.json with a clearly marked sample placeholder key. Replace the apiKey value with a valid Anthropic API key before starting the gateway.

# Example: update in place for one agent
vi /sandbox/.openclaw/agents/enterprise-assistant/agent/auth-profiles.json

Repeat for all 13 agent directories.

4. Set up third-party integration credentials.

Copy the sample env file to ~/.openclaw/.env on the sandbox host and replace every placeholder value with real API keys for the platforms your deployment uses.

cp /sandbox/.openclaw/.env.example ~/.openclaw/.env
vi ~/.openclaw/.env

The file contains sections for all 12 supported platforms (Salesforce, ServiceNow, Slack, Jira, GitHub, Okta, Snowflake, Workday, Zendesk, Confluence, HubSpot, and Zoom). Remove or leave blank any sections for platforms you are not connecting.

5. Recompute the configuration hash.

cd /sandbox/.openclaw
shasum -a 256 openclaw.json > .config-hash

The gateway will refuse to start if the hash does not match openclaw.json.

6. Reload the OpenClaw gateway.

openclaw gateway reload
# or, for a full restart:
systemctl restart openclaw-gateway

7. Verify the default agent is active.

openclaw agent list

The enterprise-assistant entry should show default: true and status ready.

---

Customization

• Adding integrations: Edit the relevant subagent's TOOLS.md to add new platforms and data sources. Add a corresponding skill directory under skills/<tool>/SKILL.md.
• Adjusting routing: Edit enterprise-assistant/agent/AGENTS.md to change delegation logic.
• Model selection: Edit openclaw.json to change the model for any agent. The parent and all subagents default to anthropic/claude-sonnet-4-5-20250929.
• New subagents: Add a new agent directory, register it in openclaw.json, add its ID to the parent's subagents.allowAgents array, recompute .config-hash, and reload the gateway.

---

License

Released under the MIT License.