Agent infrastructure— Forjio family, one runtime

Run agents that show up for you.

Discord, REST, CLI on top of one multi-agent runtime — shared memory, per-tenant tools, full tracing. Built for autonomous operations, not chatbots.

Outputs land where you live. Tools you bring. State you trust.

Start free See pricing

Free tier — BYO infra + Anthropic. No credit card.

Discord
REST
CLI

Catentio · run history

discord

hachimi

4.2s

list shopee orders delivered this week

web_fetch · search_memory

completed

tora

11.8s

audit nginx config on staging

ssh · read_file

completed

kaneko

7.1s

pay this month electricity bill

browser · pay_bill

awaiting OTP

One runtime. All agents, all transports, one trace per run.

Multi-agent runtime

one process, many personas

Per-tenant tools

BYO tokens + outputs

How it works

One pipeline, end to end.

One inbound message, one trace, one output destination. The same path whether the request comes from Discord, Slack, a REST call, or the CLI.

01
A message arrives
Discord, Slack, REST, CLI — the transport doesn't matter. The classifier picks the right agent + work mode.
02
The agent works
Memory recall, tool calls, sub-agent spawns. Every event lands in the trace; cost is tallied per run.
03
Output ships
Code goes to GitHub via PR. Files land in Drive, Spaces, or a portal download URL. Your destinations, your way.

Platform

What ships in the box.

Six things every Catentio workspace gets from day one — memory, tools, tracing, skills, safety, and outputs — wired into one runtime, same on every transport.

Memory that survives the session

Cold memory in Postgres + pgvector with daily conversation aggregation. Agents recall facts months later, semantically.

Per-tenant tool surface

YOLO baseline toolset, agent-specific allowlist, BYO tokens for everything but Anthropic. No shared service accounts.

Tracing baked in

Every tool call, every model call, every state change emits an event. Run IDs propagate across transports + sub-agents.

Skill registry

Shared SKILL.md packages installable per-agent. shadcn, ffmpeg, Notion, copywriting — load on demand, swap with a PR.

Admin OTP for sharp edges

Memory purge, secret rotate, agent kill — fresh email OTP every time. Compromised Discord ≠ compromised platform.

Output destinations you control

Code → GitHub PR or push, your repos. Files → Drive folder, Spaces, or a portal download URL. Per-task overrides.

Transports

Where the agent meets you.

Same agent, same memory, every channel. The classifier routes; the runtime executes.

Discord
live
Slack
planned
Telegram
planned
WhatsApp
planned
CLI
live
REST
live
Meet
live (Chrome+Meet, internal)

Pricing

Pick caps first. Pick infra second.

Tier sets your platform capability — workspaces, agent count, memory size, runs/day. Each paid tier has BYO and Hosted variants so you choose whether to run the boxes yourself or have catentio run them.

Free

Tinker. BYO everything.

1 workspace · 1 user
3 agents · 5 tools · 5 skills
100 runs / day
500 cold-memory entries

Self-host

Rp 0

no card

Starter

Solo / small team.

1 workspace · 3 users
20 agents · 50 tools · 50 skills
5,000 runs / day
50,000 memory entries
Email support

BYO infra

Rp 290.000

/mo

Hosted

Rp 1.490.000

/mo

Popular

Pro

Power users / agencies.

5 workspaces · 15 users
100 agents · 200 tools / skills
Unlimited runs
500k memory entries · 25 heartbeats
Email + 24h response
Managed Anthropic available

BYO infra

Rp 990.000

/mo

Hosted

Rp 3.490.000

/mo · Rp 2M Anthropic credit included

Scale

Multi-region · mTLS · SLA.

Unlimited everything
Dedicated agent runner
Multi-region + mTLS
Custom token markup tiers

Custom

Quote

Start free Talk to us about Scale

Built-in agents, tools, and skills don't count against caps — only your custom-authored ones do. Forjio-internal workspaces are unlimited (you're billed nothing).

Indonesian customers pay in IDR (QRIS · VA · e-wallet · card) via Plugipay. International customers pay in USD via PayPal — Midtrans doesn't process USD.

Compare

Where Catentio fits.

You get the runtime, the memory, the tracing — and the transports — without gluing five libraries together yourself.

Capability	Catentio	LangChain	CrewAI	DIY scripts
One runtime, many transports				glue code
Multi-agent + classifier router		partial
Persistent shared memory	pgvector + Redis	BYO store	BYO store	BYO
Per-run tracing in your DB		LangSmith	partial
Per-tenant BYO tokens				BYO
Output-destination router				BYO
Auth-gated portal + admin OTP				BYO
Self-hostable, no vendor lock-in

One login

Sibling, not part of the suite.

Catentio is its own thing. It logs in via Huudis, the same identity layer the rest of the family uses — that's it. No bundle, no upsell.

You are here

Catentio

Agent infrastructure

Plugipay

Payment infra for Indonesian sellers

LinkSnap

Branded short URLs + QR analytics

Storlaunch

Storefront-first commerce platform

Auth via Huudis — one login across the family.

FAQ

Common questions.

How do tiers and variants relate?
Tier sets your platform capability — workspaces, agent count, runs/day, memory size, support level. Variant decides who runs the boxes: BYO (you do, on your own VPS) or Hosted (catentio runs them). Free is BYO-only; Starter and Pro have BYO + Hosted; Scale is custom.
What does the Free tier actually include?
Self-hosted on your own VPS / Postgres / Redis / storage. 1 workspace · 1 user · 3 custom agents · 5 tools · 5 skills · 100 runs / day · 500 cold-memory entries. Built-in agents / tools / skills (anything shipped with the catentio repo) NEVER count against caps — only your custom-authored ones do. No card required.
Whose Anthropic tokens does it use?
On Free, Starter, and Pro BYO: yours. Bring an API key or an Anthropic Max subscription — we never see the tokens. On Pro Hosted only: managed Anthropic is included. We mint a per-customer workspace under our forjio org and bill tokens at cost × 1.20 against a Rp 2,000,000 monthly credit bundle; overage rolls into next cycle's invoice.
How does memory work across sessions?
Three layers: thread history in Redis (7-day TTL), cold memory in Postgres + pgvector with semantic recall, and daily conversation aggregation that auto-indexes long-term context. Agents recall details from months ago without you re-explaining.
Can I bring my own agents / tools / skills?
Yes. Today the runtime ships create-agent and create-tool over the control-plane REST API, and install-skill copies skill bundles into the workspace at runtime. A draft-PR review surface is on the roadmap so code-path changes can be reviewed before they land; for now changes apply against your runtime database directly and are versioned + auditable per workspace.
What happens if I lose access to my Discord account?
Sensitive operations (purge memory, rotate secret, kill agent) require a fresh OTP sent to your email — every time, not just at session start. A compromised transport can't escalate.

Ready when you are.

Free tier ships with everything you need to prove fit on your own infra. Upgrade to Starter or Pro when caps start pinching.

Start free Talk to a human