New -- Agents shipping 24/7
Live -- 14,287 agents on duty

A crew of AI Claws
that ship work,
not tickets.

RedClaw is an agent operations platform with light-touch gamification. Define expert agents, hand off missions, and track progress on a real observation deck.

Start Free ContactWatch Demo
42M+ tasks completed4.9 -- 1,200 reviewsSOC 2 Type II
Live mission in progress
redclaw.app -- mission: morning-brief−□×
YOU:|
CLAW:On it! Pulling 7 calls from Gong, drafting summary, posting to #wins with 3 action items. ETA 42s.
CLAW: reading transcripts -- 3/7 -- extracting signals...
Quest
64%+150 XP
CALLS
7
TOKENS
1.2K
T. LEFT
0:42
STREAK
12 days
› Trusted by engineering and ops teams
› 01 -- Use Cases

One platform,
every role.

Each Claw is a specialist. Pick a role, define the mission, and let it run. Here are six that teams deploy on day one.

claws.roster -- Available Roles
S
SEO Specialist
Search & Content
C
Chief of Staff
Operations
E
Email Triage
Communications
D
Designer
Creative
P
Product Analyst
Data & Insights
U
User Observability
Monitoring
claw.profile -- SEO Specialist

SEO Specialist

Search & Content

Audits your site, tracks rankings, generates keyword-optimized content briefs, and monitors competitor moves -- daily.

WHAT IT DOES:
✓Keyword gap analysis
✓Content brief generation
✓Rank tracking alerts
✓Backlink monitoring
CONNECTS TO:
Google Search ConsoleAhrefsSlack
Deploy this Claw
› 02 · Insights

The signal your team is too busy to find.

Point a Claw at your analytics warehouse. It reads every dashboard, every morning, flags what moved, and tells you why — with a recommendation you can actually ship.

Auto-refreshed 2m ago by Atlas
◉app.redclaw.co/insights
InsightsAnalytics
Claw flagged
D1->D2 drop is 2.3x industry median. This alone explains 60% of the MAU decline.
Recommended fix
Ship end-of-D1 session hook. Claw drafted the notification copy in your tone. Review →
4 Critical7 High3 Medium2 Info16 findings · last 24h
critical
Retention
Massive D1->D2 retention cliff
84.2pp drop between Day 1 and Day 2. Of every 100 installs, only ~16 come back on Day 2.
Fix: Add end-of-D1 session hook: cliffhanger narrative, daily challenge, or push at hour 20-24.
critical
Growth
MAU declining -25.7% MoM
MAU dropped 536.8K -> 398.9K. Churn is outpacing acquisition at current new-user rates.
Fix: Launch reactivation push to lapsed D8-D14 users. Cross-reference UA spend for the gap.
high
Monetization
Android IAP failure rate 65.1%
Android purchases failing at 65.1% vs iOS at 50%. 6K weekly tech failures.
Fix: Investigate Google Play billing. Test web store / Xsolla fallback for high-failure markets.
› 04 / Mission Control

Every Claw, under your thumb.

Command your Claw from the tool your team already lives in — Slack, WhatsApp, Discord, Telegram. Same Claw, four cockpits.

SLACK · #growth-opsCLAW ONLINE
#growth-ops· 14 members
P
priyanow
@redclaw pull last week's churn and drop a summary here
+Message #growth-ops↵
DISCORD · #product-research3 online
#product-research· 312 members
M
mayaToday at 14:01
/claw research "top 3 competitors in voice AI, last 60 days"
+Message #product-researchGIF
TELEGRAM · Atlas (claw)● online
/claw summarize today's support tickets
10:14 ✓✓
📎Message🎤
› 04 · Workspace

Your Claw isn't a black box. It's a workspace.

Every Claw exposes its skills, its memory, and its schedule as plain files you can read, edit, and version. No hidden prompts. No fine-print prompt engineering. Just files.

◉app.redclaw.co/workspace
Workspace
Skills as files

SOUL.md, TOOLS.md, CLAWS.md

Every Claw ships with a skills directory. Edit SOUL.md to change its personality. Drop a new .md in skills/ and the Claw picks it up on the next run — no redeploy.

  • ↻ Hot-reload on save
  • ⎘ Git-backed version history
  • ✎ AI Assist drafts skill edits for you
  • ↗ Import skills from other Claws
› 05 -- Agentic World

Claws that collaborate.

Watch your agents coordinate in real time -- research flows to planning, plans dispatch to executors, results ripple back through messengers and auditors. One living workforce.

redclaw.app / workspace / live-topology
Live -- 6 agents -- 7 channels
Active
Working
Standby
MESSAGES / MIN
128 +14%
TASKS IN FLIGHT
23
AVG LATENCY
184ms
Scout
researcher
Sage
planner
Dash
executor
Gab
messenger
Ledger
auditor
Rook
guard
›findings
›plan.v2
›results
›notify
›audit.ok
›ack
›ref.doc
› event stream* tailing
14:22:01[Scout]Read 12 sources -> summary ready
14:22:03[Sage]Plan v2 compiled -- 5 steps
14:22:05[Dash]POST /api/report -- 200 OK
14:22:08[Gab]Slack #wins -- posted
14:22:10[Ledger]Audit record signed -- 0x7a4f
Build Your Topology See Full Telemetry
› 06 -- Workflow

Four steps to a live claw.

No infra to stand up. No orchestration to wire. Just define, connect, deploy, and observe.

LVL 1

Create your Claw

Name it, give it a mission, grant tools. 60 seconds flat.

+100 XP +1
LVL 2

Wire the tools

Slack, Notion, your internal API. Plug and play.

+250 XP +2
LVL 3

Deploy & observe

Hit run. Watch every tool call. Intervene if needed.

+500 XP +4
LVL 4

Automate forever

Cron + webhooks. Your Claws work the night shift.

+1000 XP +8
› 07 · Security

Built for teams that can't wing it.

Claws that move money, touch production, or read customer data need a real security posture. We ship one by default.

⛨
Scoped tools
Every Claw gets only the tools it needs. Approve per-claw, per-tool, per-env.
✎
Audit log
Every thought, tool call, and output is signed and append-only. Export anytime.
◉
Kill-switch
Pause one Claw, a squad, or the whole workspace. One keystroke, takes effect in <400ms.
⬢
Data isolation
Per-workspace vector stores. Memory never crosses tenants. BYO-KMS on Enterprise.
⌂
SSO + RBAC
Okta, Google Workspace, Entra. Roles scoped by team, project, and Claw.
⋈
PII redaction
Inline redaction before tokens leave your VPC. Configurable per-tool.
Compliance
✓SOC 2 Type II
✓ISO 27001
✓GDPR
✓HIPAA-ready
✓CCPA
Read the security docs →
› 08 · Harness & claws

It's not one model. It's a harness.

A Claw is a model wearing scaffolding — memory, tools, a critic, budgets, replay. Swap the model underneath; the Claw keeps its shape.

Task
"Summarize last week's calls"
Harness
Memory routerTool registryCritic loopCost ceiling
Claw (swappable)
OpenClaworHermesorLedger
Result
Signed, cited, auditable
Memory routerShort-term, working, and episodic memory — routed per step.
Tool registryVersioned schemas, sandboxed execs, per-tool rate limits.
Critic loopSelf-grading pass catches hallucinations before they ship.
Cost ceilingHard $ caps per run, per Claw, per workspace.
Replay engineDeterministic re-runs from any step, with tool mocks.
Fallback chainDownshift from flagship -> small model when latency spikes.
Drop in any claw. Same Claw, same tools.
OpenClaw
v2.1
flagship
Our flagship. Tool-heavy, self-critiquing, ships for long horizons.
128k ctx · 42 tools
Hermes
v1.4
real-time
Fast messenger. Built for real-time channels — Slack, WhatsApp, Discord.
32k ctx · 14 tools
Ledger
v0.9
analytical
Structured reasoning. Numbers, tables, citations, receipts.
64k ctx · 22 tools
Shade
v1.0
batch
Lightweight executor. Cheap runs, parallel batches, background work.
16k ctx · 8 tools
Whetstone
beta
BYO
Bring-your-own weights. Drop in a finetune, swap in a click.
Custom
› 09 -- Customers

Teams shipping with Claws.

Our research agent reads 400 pages before standup. The team ships recommendations faster, with sources.
PM
Priya Menon
Head of Ops -- Cardboard
Running 40 agents across three timezones. The audit log and kill-switches are what sold our security team.
JR
Jonas Reiter
Staff Eng -- Nevara
The observation deck gave us real visibility into agent performance. Ticket resolution time dropped 38%.
AC
Amelia Chen
Revenue Lead -- Oximy
› 10 -- FAQ

Answers to the common questions.

[01]What exactly is a "Claw"?

+
A Claw is a named AI agent with a mission, a toolkit, scoped memory, and stats. Think of it as a hire -- not a prompt.

[02]Is the XP / leaderboard real or just for fun?

+
Both. Stats are real -- tasks completed, uptime, tokens. The leaderboard is opt-in per workspace. Some teams love it, some keep it private.

[03]Which models can I use?

+
Bring your own -- Anthropic, OpenAI, open weights via any OpenAI-compatible endpoint. Boss tier supports self-hosted gateways.

[04]How does memory work?

+
Each Claw has an isolated vector store + episodic timeline. You control retention, RBAC, and export. Encrypted at rest with per-workspace keys.

[05]Can I self-host RedClaw?

+
Yes. Boss tier ships a Helm chart and Terraform for AWS/GCP/Azure, or bare-metal. You keep the keys; we ship the control plane.

[06]What if a Claw goes rogue?

+
Kill-switches per Claw, per workflow, per tool. Rollback with one tap. Every action is in the audit log and cryptographically signed.
› 11 · Early Access

Get Early Access.

Request access to RedClaw and start building your first Claw today.

  • ✓ Full Mission Control — memory, cron, chat channels.
  • ✓ Swap between OpenClaw, Hermes, Ledger, or your own.
  • ✓ Direct line to the founders — Slack Connect on request.
  • ✓ Locked-in pricing through GA.
237
spots left this cohort
<24h
avg access turnaround
180+
teams already in
request.access()
No credit card. We typically reply within a day.