Fleet -- Agent Management

Deploy, manage, and monitor AI agents with built-in trust controls. Fleet is VeriSwarm's service layer for operating agents with Gate scoring, Guard security, Passport identity, Vault audit, and Cortex runtime already wired in.

Plan requirement: Fleet features are available on all plans. Some capabilities (Guard scanning, Passport verification, Vault audit) require the corresponding plan tier.

---

Agent Templates

Fleet includes 6 built-in templates for common agent workflows. Each template comes with pre-configured channels, compliance settings, pricing guidance, and a SOUL personality definition.

Template	Use Case	Key Integrations
Generalist	General-purpose assistant	Slack, Gmail, Google Calendar
Healthcare	Patient-facing clinical support	FHIR, Insurance Verification
Real Estate	Property search and lead management	kvCORE, Zillow, Follow Up Boss
Support	Customer service and ticketing	Zendesk, Freshdesk, Intercom
Accounting	Bookkeeping and financial ops	QuickBooks, Xero, Stripe, Expensify
Legal	Case management and document review	Clio, MyCase, DocuSign

Browse templates from the Agents page in the dashboard. Each template shows a description, included integrations, compliance presets, and a deploy button.

---

Custom Templates

Build your own agent template when the built-ins do not fit.

SOUL Editor

The SOUL defines the agent's personality, boundaries, and behavioral rules. Write it in markdown with structured sections:

• Identity -- who the agent is and how it introduces itself
• Boundaries -- what the agent will and will not do
• Tone -- communication style and formality level
• Compliance -- industry-specific rules (HIPAA, PCI-DSS, etc.)
• Escalation -- when to hand off to a human

SOUL content is security-scanned on create and update. The scanner checks for prompt injection patterns, bypass attempts, and content that conflicts with the agent's compliance profile.

Config

The config.json defines channels, pricing, LLM configuration, and integration requirements. Custom templates are tenant-scoped -- they are only visible within your workspace.

Managing Custom Templates

• Create: POST /v1/agents/marketplace/custom-templates
• Update: PUT /v1/agents/marketplace/custom-templates/{id}
• Delete: DELETE /v1/agents/marketplace/custom-templates/{id}
• Export: GET /v1/agents/marketplace/custom-templates/{id}/export
• Import: POST /v1/agents/marketplace/custom-templates/import (SOUL is re-scanned on import)

---

Deploy Wizard

Deploy an agent in five steps:

1. Template -- select a built-in or custom template. The template pre-fills defaults for all subsequent steps.
2. Pricing -- configure the agent's billing model (free, subscription, per-query, or custom).
3. LLM Config -- select primary and fallback LLM providers, choose models for simple and complex queries, set temperature and token limits. See Cortex for routing details.
4. Business Info -- set the agent's display name, slug, description, and owner details.
5. Channels -- enable communication channels (widget chat, API, Slack, Telegram, etc.) and configure per-channel settings.

Click Deploy to create the agent. The agent immediately appears in your workspace with Gate scoring active.

---

Agent Manage Page

After deployment, the manage page provides eight tabs for operating the agent.

Settings

• Display name, slug, description, and avatar
• LLM configuration (provider, model, temperature, max tokens, fallback chain)
• Widget embed snippet and widget token management
• Allowed domains for widget embedding
• Agent visibility (public profile on/off)

Integrations

Connect external services on a per-agent basis. Each agent can have its own set of active integrations, independent of workspace-level connections.

• OAuth integrations (Google Calendar, Gmail, HubSpot, Slack, etc.) -- click Connect and authorize
• API key integrations (Stripe, Twilio, Zendesk, Jira, etc.) -- enter credentials and save
• 29 integrations available across scheduling, email, CRM, documents, messaging, support, payments, legal, real estate, and healthcare categories

See Agent Integrations for the full list and setup instructions.

Knowledge Base

Upload documents or crawl URLs to build the agent's retrieval corpus.

• File upload -- PDF, DOCX, TXT, MD, CSV. Files are chunked, embedded, and indexed for retrieval.
• URL crawling -- provide a URL and the crawler extracts page content. Useful for FAQ pages, documentation sites, and product pages.
• Document management -- view, search, and delete uploaded documents. Re-index after changes.
• RAG retrieval -- when the agent receives a query, relevant chunks are retrieved from the knowledge base and injected into the LLM context. The agent's response is grounded in your content.

Chat Test

Send test messages to the agent directly from the dashboard. Useful for verifying LLM config, knowledge base accuracy, and SOUL behavior before going live.

Conversation Logs

Browse all conversations the agent has had, across all channels (widget, API, Slack, etc.). Each log shows:

• Timestamp and channel
• Full message history (user and agent turns)
• Guard findings (PII detections, injection attempts)
• Cost per conversation (token usage and model costs)
• User identity (for secure-mode widget sessions)

Filter by date range, channel, or search by message content.

Vault

View the agent's immutable audit trail (requires Max plan with Vault enabled):

• Hash-chained ledger entries for every conversation, decision, and integration call
• Chain integrity verification
• Export to JSON or CSV for compliance

Passport

Manage the agent's verified identity (requires Pro+ plan with Passport enabled):

• Identity verification status
• Signed capability manifests
• Active delegations (user-to-agent permission grants)

Security

Guard security controls (requires Max plan with Guard enabled):

• Security findings (credential leaks, PII exposure, injection attempts)
• Tool policies (allowlist, MCP restrictions, content filters)
• Kill switch status

---

Widget Embed

Get the embed snippet from the Settings tab on the manage page. See Widget Chat for full configuration details, authentication modes, and security settings.

---

Agent API Keys

Each agent can have its own API key (agta_...) for self-service operations:

• Issue: generated automatically on agent creation
• Rotate: POST /v1/public/agents/{agent_id}/api-keys/rotate -- generates a new key and invalidates the old one
• Revoke: POST /v1/public/agents/{agent_id}/api-keys/{key_id}/revoke -- permanently deactivates a key
• List: GET /v1/public/agents/{agent_id}/api-keys -- view active keys

Agent keys are used for portable credential issuance (POST /v1/credentials/issue) and self-service score retrieval (GET /v1/agents/me/scores).

---

Kill Switch

Emergency stop for a compromised or misbehaving agent. Activating the kill switch:

1. Blocks all trust decision checks for the agent
2. Denies credential issuance
3. Creates an immutable Vault ledger entry (if Vault is enabled)
4. Sends a danger-level notification to the workspace

Activate from the Security tab on the manage page, or via the API:

curl -X POST https://api.veriswarm.ai/v1/suite/guard/kill/agt_123 \
  -H "x-account-access-token: vsa_your_token"

The kill switch is reversible. Only accounts with guard.killswitch.write permission can activate or reverse it.

---

Delete Agent

Deleting an agent removes:

• The agent profile and all configuration
• Knowledge base documents and embeddings
• Conversation logs
• Guard findings and policies scoped to the agent
• Passport manifests and delegations
• Widget tokens

Vault ledger entries are retained (they are immutable). Deletion is permanent and cannot be undone.

---

Monitoring

Conversation Logs

Real-time and historical logs across all channels. Filter, search, and drill into individual conversations.

Vault Audit Trail

Immutable record of every action taken by or on behalf of the agent. Useful for compliance reviews, incident investigations, and customer disputes.

Cost Tracking

Per-agent cost breakdown by LLM provider and model. Tracks token usage, estimated cost per conversation, and daily/weekly/monthly aggregates. Set budget limits via Cortex budget enforcement.

---

Related Docs

• Widget Chat -- embeddable chat widget configuration
• Cortex -- Runtime -- LLM routing, cost tracking, and budget enforcement
• Agent Integrations -- connecting external services
• Guard -- security scanning, PII tokenization, and kill switch
• Passport -- identity verification and delegations
• Vault -- immutable audit ledger
• API Reference -- full endpoint details