Opportunities › Agent orchestration and credential management for teams › Proposed app
Sluice
Run team AI agents with keys kept out of reach and every action logged
- Score
- 46
- confidence 49
- Evidence
- 9
- items on the page
- Rivals
- 3
- named in sources
- First version
- 60 d
- estimate
- Type
- Desktop and mobile
- PWA: on computers and phones alike
- Market since
- 2025
- New market · AI estimate
ScreensConcept
Desktop and mobile: Teams configure agents and policies at desks, but must approve risky actions immediately on phones as agents run throughout the workday. Sample screens on a computer and a phone, in the proposed brand. Their numbers are sample data.
Agents
All gateways online6 running
About this appAI proposal
Sluice is a control console for teams that run autonomous AI agents. Each agent works in its own isolated VM. Credentials never enter the agent: a gateway adds them to a request only after a policy check, so a prompt-injected agent has no key to leak.
Admins set organization-wide rules on what each agent may reach, who on the team can use it, and which actions need approval. Rules are enforced at the network layer, outside the agent and the model. Every action is logged with the agent, the person and the credential used.
The gateway runs on your own infrastructure and the core is open source. The web app is the place where security and operations staff review, approve and audit. It is meant for teams that now build their own agent harness again for each use case.
- Tagline
- Agents do the work. Keys stay behind the gate.
- Voice
- PlainPreciseCalm
- Why this name
- A sluice is a gate that controls what flows through. Agents get through only what policy lets pass. The word is short and easy to say.
In the first version
Members say which features they would use. It tells PWAS what people want; it changes no score.
Credentials injected at the gateway
Agents hold no keys. The gateway adds a credential to an outbound call only after policy allows it, so keys are not in memory, files or session logs.
No member yet
One VM per agent
Each agent runs isolated from others and from the host, so a compromised or confused agent cannot reach another team's tools or data.
No member yet
Policies outside the model
Allow, deny or require approval by service, endpoint and action. Rules apply at the network layer, so prompt injection cannot talk its way past them.
No member yet
Team roles and shared agents
Give sales ops, engineering and admin their own agents and permissions. Invite colleagues, assign who can run, edit or approve.
Brand identityAI proposal
Check the name against trademark registers before using it.
Palette
#14585CDeep Teal · primary#F4F1EAPaper · surface#1F2A2EInk · text#D98A1FAmber · accent#3B6EA8Slate Blue · data#B23A2EBrick · deny and alerts
Typography
Aa
Sluice
PricingAI proposal
The material gives no competitor prices. Tiers follow the open-core idea in the opportunity; amounts are left open and are a starting point to test with early teams.
Community
Free
- Self-hosted open-source gateway
- Single team, few agents
- Basic audit log
Team
Contact us
- Roles and shared agents
- Approval queue and mobile alerts
- Policy templates
- Log export
Enterprise
Contact us
- Multi-tenant administration
- Advanced policy and audit reports
- SSO
- Priority support
What it could be worthEstimate
What a software business making this much sells for today, from real deals. Pick a monthly revenue.
If it makes
Worth about
$194,000
Sale range $132,000 to $263,000
$120,000 a year in revenue
From 87 software businesses listed and sold on Empire Flippers in the last year, revenue vetted by the broker: they sell for about 19× their monthly revenue, most between 13× and 26×.
Measured on profit, the middle deal is 32× monthly net profit: about 2.7 years of profit.
An estimate, not an offer or financial advice. A real price depends on profit, growth, churn, how much of the work is the owner's, and the buyer.
How it would growAI proposal
How the product would reach new customers through its own use, and how the first ones would be found. A plan to test, not a result.
- Built-in loop
- No loop in the product. Reached through open-source adoption, content on agent security and direct sales. Self-hosted use inside a company is not visible to outsiders, so no loop is assumed.
- Do they become customers?
- Test whether engineers who try the free gateway bring in a security or ops buyer who pays for roles, approvals and audit. Colleagues invited inside one company are the only built-in spread.
- First customers
- Engineering or platform leads at software companies of 20 to 200 people already running autonomous agents for sales operations or engineering and writing their own harness.
- Where to reach them
- Hacker News and open-source repo: Publish the gateway, a write-up on keeping keys away from agents, and a demo of a blocked prompt-injection attack.
- DevOps and security communities: Offer setup help in agent and platform-engineering forums and chats to teams that describe credential problems.
- Direct outreach: Contact security leads at companies posting about agent rollouts and offer a free pilot.
- First milestone
- Five teams run the gateway on real workflows, and at least two agree to pay for the team tier within 60 days.
Why people already pay for this
- VerifiedHacker News post received 88 points and 37 comments
- Public claimAgents store credentials insecurely in memory, files, and session logs as plain text
- Public claimAgents can be manipulated to expose API keys through prompt injection
- Public claimTeams repeatedly build custom agent harnesses with missing credential and permission management
- Public claim
Tech stack
Every proposal builds on the same PWA base, installable from the browser on any phone or computer. Dashed boxes are what this app adds.
- AppNext.js and TypeScript
- PWAManifest and service worker
- UITailwind CSS
- Data and authSupabase Postgres
- PaymentsStripe
- EmailResend
- HostingVercel
- This app · Agent gatewaySelf-hosted proxy service in Go or RustMust sit in the network path, check policy and inject credentials outside the agent and the LLM.
- This app · VM isolationMicroVM runtimeGives each agent its own isolated machine on the customer's infrastructure.
RisksAI inference
- TechnicalIsolation and gateway injection must hold against hostile prompts. One bypass, such as a key leaking in a response body, would destroy trust. Needs independent security review before launch.
- CommercialOpen-source core may satisfy small teams, leaving paying demand only in larger firms with long sales cycles. The paid line between core and extensions must be tested.
- PlatformAgent frameworks change fast and some hosted agents cannot be wrapped. Supporting a few frameworks well may limit the reachable market.
- RegulatoryRegulated buyers may ask for compliance attestations and data-handling guarantees that a small team cannot provide early.
- AIPolicies on endpoints and actions can miss harmful behavior that looks allowed, such as an agent sending real data to an approved service.
Everything marked AI proposal or AI inference was written by a model from the evidence on the opportunity page. The proposed name must be checked against trademark registers before it is used.