Praes
Built for OpenClaw.OpenClaw

The cockpit for
your agent.

Get startedDocumentation
Activity · support-agent
live
Status / ModelStartedCost
Success · gpt-5.3-chat
14s ago$0.0124
Running · claude-sonnet-4.6
now$0.0041
Error · mercury-2
2m ago$0.0009
Success · minimax-m2.5
3m ago$0.0063
Success · gpt-5.3-codex
6m ago$0.0018
Running · mercury-2
8m ago$0.0029
Success · claude-opus-4.6
12m ago$0.0142

Watch it in action

A clean browser frame ready for your product walkthrough.

praes.app

What you get

Everything you need to observe, debug, and improve your agent in one calm dashboard.

ActivityMemorySoulCostsTools
Run visibility

Trace every run.

Inspect status, model, retries, tools, and full event timelines in one glance.

Status / ModelStartedCost
Successgpt-5.3-chat
14s ago$0.0124
Runningclaude-sonnet-4.6
now$0.0041
Queuedminimax-m2.5
22s ago$0.0019
Errormercury-2
2m ago$0.0009
97.4%
Success rate
1.8s
Median latency
$0.0091
Median run cost
Run timelineLive events
14:01:03
Intent parsed and routing selected
14:01:04
semantic_search returned 6 matches
14:01:05
refund_policy tool validated eligibility
14:01:06
Final response passed SOUL checks
14:01:07
memory.md context injected into prompt
14:01:08
Streaming reply delivered to customer

Active chain

search → refund_policy → response

Guardrail state

SOUL pass • 0 warnings

Memory workflows

Edit memory.md like a product.

Review, tag, and evolve memory safely with clear provenance and fast search.

policy/refunds.md

Updated by support-agent • 4m ago

Pinned

brand-voice.md

Reviewed by ops-lead • 17m ago

Reviewed

escalation-rules.md

Synced from memory.md • 53m ago

Synced
Cost clarity

See live costs per run.

Understand token and tool spend as runs happen, then optimize with confidence.

gpt-5.3-chat$182.41
claude-sonnet-4.6$109.16
minimax-m2.5$47.88
Soul guardrails

Keep behavior aligned.

Make SOUL.md edits safer with visible principles, do/don't checks, and escalation rules.

  • Purpose and tone checks before every final response
  • Escalation policy surfaced when confidence drops
  • Tool permissions validated before execution
Tool reliability

Know where tools break.

Track call volume and error rates per tool so regressions are obvious before they escalate.

ToolCallsError rate
semantic_search1,2480.4%
ticket_lookup9320.9%
web_fetch2871.8%
refund_policy6110.2%

FAQ

Straight answers to the questions teams ask before pairing Praes.

Ready to run your agent with clarity?

Connect once, then monitor every run, memory decision, and cost signal from one calm dashboard.

Get startedCommunity