Product

A control plane for your AI workforce.

Not a chat box. Not a vibe-coded prompt. A working operational system: deploy agents, give them tools, watch them work, swap models, audit every action.

Dashboard

See your AI workforce at a glance.

The Workforce card shows every active agent, what each is doing right now, how many tasks each has completed today, and what they cost. Click any agent to see session history, recent tool calls, and per-token spend.

  • Per-agent token + cost attribution
  • Live activity feed across all agents
  • Health checks: failing tasks, overdue items
  • Channel status (Discord, Slack, Email, Web)
app.inhouseagents.ai

Dashboard

What needs attention, what's been happening, system health.

Workforce

8 agents · today
Tasks
142
Active
8
Cost
$0.00
SDR Booked meeting with Acme Corp 2m ago
Support Resolved ticket #482 4m ago
Designer Drafted UI spec for new dashboard 6m ago
Billing Recovered failed payment ($4,900) 12m ago
Needs attention

All clear — nothing failing or overdue.

Channels

5/5 connected · Discord, Slack, Email, Web, Telegram

Roles & Teams

Pre-built agents. Pre-wired teams.

8 starter roles cover the most common back-office work. Each agent ships with a system prompt, the right tools, and a heartbeat schedule for recurring work. Team Packs bundle 3+ agents with delegation pre-configured — install a Product Team or Customer Lifecycle Team in one click.

  • SDR, Support, Billing, EA, Research, PM, Designer, TPM
  • Edit any prompt, tool list, or model after install
  • Team Packs wire delegation graphs automatically

Roles

📞
SDR Installed

Prospects, researches, books meetings

💬
Support Agent Installed

First-response triage + KB lookups

💰
Billing Coordinator Installed

Invoicing, dunning, recovery

📅
Executive Assistant

Scheduling + inbox triage

🔍
Research Analyst

Briefs, memos, charted findings

🧭
Product Manager

Coordinates Designer + TPM

Capability Slots

Agents declare needs. You bind tools.

An SDR needs a CRM, an email sender, and a lead source. Inhouse Agents abstracts these as capability slots. Bring HubSpot, or Salesforce, or Pipedrive — bind it once and every agent that needs a CRM uses yours.

No hard-coded vendor lock-in. Switch CRMs in production without rewriting agents.

See which capabilities each agent needs →

Capability Slots

Bind agent capabilities to whatever tools you actually use.

🎯
CRM
3 agents need this
HubSpot
📧
Email send
3 agents need this
Gmail API
🔎
Lead source
3 agents need this
Apollo
📅
Calendar
3 agents need this
Google Calendar
🎫
Helpdesk
3 agents need this
— unbound
💳
Billing
3 agents need this
Stripe
API Wrappers

Got a custom internal tool? Wire it in 5 minutes.

Paste the API docs (OpenAPI spec, Postman collection, or just markdown describing the endpoints). Inhouse parses the spec, generates the tool bindings, and your agents start using it on the next loop.

When the API drifts — field renames, new required headers, deprecated endpoints — agents adapt automatically. Most schema drift is handled without a redeploy or an engineering ticket.

  • Works with your homegrown CRM, billing, ticketing, or any REST API
  • Self-heals on schema drift via response-body inspection + retry
  • OAuth, bearer, basic, or API-key auth — all native
  • Rate-limit aware; respects retry-after headers
app.inhouseagents.ai/packs/custom
API Docs (paste link or markdown)
https://docs.acme-billing.internal/openapi.yaml
Wrapper generated · 14 endpoints detected
POST /invoices · create_invoice
GET /invoices/:id · get_invoice
POST /payments/charge · charge_card
POST /refunds · issue_refund
+10 more
Bound to Billing Coordinator · ready to use

Production-grade infrastructure.

🔌

Webhook triggers

Wire any external system to fire an agent. Stripe payment failed → Billing agent triages. Linear bug filed → TPM analyzes. Templated prompts render webhook payloads as natural-language briefs.

POST /trigger/billing-failed
Authorization: Bearer <secret>
{"customer": "acme", "amount": 4900}
📊

Observability

Every action is logged: tools called, tokens used, latency, success/fail. Per-agent rollups and per-task drill-down. Cost attribution down to the request.

Streams from a JSONL metrics log — your audit trail is grep-able.

🔐

BYO-app OAuth

Use your Google/Microsoft/Salesforce/Slack/HubSpot OAuth apps, not ours. Tokens encrypted at rest with AES-256-GCM. PKCE flows. Scope minimization by default.

🤝

Agent-to-agent delegation

Product hands a UI brief to Designer, then a sprint plan to TPM, then synthesizes both. Depth-limited (3 hops max). Whitelisted per-agent. Visible in the activity log.

See it running on your stack.

Book a 15-min demo →