WorkingAgents: Feasibility and Probability of Success

An honest assessment of whether WorkingAgents can become a viable business, drawing on internal documents, market data, and the competitive landscape as of March 2026.

What WorkingAgents Is

WorkingAgents is an AI agent governance platform built on Elixir/OTP. It puts a control plane between AI agents and everything they touch – LLMs, tools, data, and communication channels. Three gateways (AI Gateway, AI Agent Gateway, MCP Gateway) provide unified routing, permission enforcement, audit trails, and guardrails.

The product positioning: “Companies built systems to manage employees – identity, permissions, accountability, audit. Enterprises will require the same governance for AI agents.”

Two products serve different entry points:

One-instance-per-customer deployment. Self-hosted. Zero data egress. Built by a solo founder based in Vietnam.

The Market Case: Strong

The tailwinds are real and measurable.

The agent wave is happening. MCP SDK downloads grew from 100,000 in November 2024 to over 8 million by April 2025, with 10,000+ active public servers. A2A launched in April 2025 with backing from every major hyperscaler. Gartner predicts 40% of enterprise applications will embed task-specific AI agents by end of 2026, up from less than 5% in 2025. The autonomous AI agent market is estimated at $8.5 billion in 2026, potentially $35-45 billion by 2030 (Deloitte).

Governance is the bottleneck. While 38% of enterprises were piloting agents by late 2025, only 11% reached production. The gap isn’t model performance – it’s durability, state management, and governance. A survey of 205 infrastructure and security leaders found 43% had no formal AI governance controls. Over-privileged AI had a 76% incident rate versus 17% for least-privilege AI. Security teams are the #1 blocker of AI adoption in enterprises.

Regulation is coming. The EU AI Act’s high-risk provisions take effect August 2026. Enterprises deploying AI agents in healthcare, finance, or HR will need documented governance frameworks, audit trails, and risk assessments. This creates mandatory demand for exactly what WorkingAgents provides.

Budget exists. Big Tech is expected to invest roughly $650 billion in AI infrastructure in 2026. McKinsey reports 88% of respondents use AI in at least one business function. 75% of companies may invest in agentic AI by end of 2026.

Verdict: The market is real, growing fast, and has a clear governance gap. This is not a solution looking for a problem.

The Product Case: Credible

WorkingAgents has a working product, not a pitch deck. The platform runs in production with 86+ MCP tools covering CRM, task management, content, scheduling, monitoring, email, WhatsApp, and knowledge management. Key technical strengths:

Capability-based access control. Permission keys are compiled into modules at build time. Each agent inherits the user’s permissions. A single is_map_key guard check gates every tool call – O(1), no allocation. This is architecturally sound, not bolted on.

Elixir/OTP foundation. Fault-tolerant supervision trees, hot code reloading, per-user process isolation, sub-millisecond message passing. The same runtime that powers WhatsApp and Discord. This is a genuine technical advantage for a system that manages long-running, stateful agent workflows.

Protocol-native. Built on MCP and A2A from day one, not retrofitted. Virtual MCP Servers provide per-user tool scoping. A2A enables cross-platform agent discovery and delegation.

Self-hosted, zero egress. Each customer runs their own instance. No shared infrastructure, no data leaving the customer’s environment. This bypasses the security review bottleneck that kills most AI vendor evaluations.

Verdict: The product is real, technically differentiated, and addresses the governance gap directly. The Elixir/OTP choice is both a strength (reliability, concurrency) and a risk (smaller talent pool, enterprise unfamiliarity).

The Competition Case: Crowded and Getting Worse

This is the biggest risk. The competitive landscape is intense and accelerating from three directions simultaneously.

Platform giants are building natively. Microsoft Copilot Studio has A2A support. Amazon Bedrock AgentCore offers secure tool access with centralized policy. Google Vertex AI Agent Builder covers governance and IAM-based agent identity. Salesforce has Agentforce. SAP has Joule. These companies can bundle agent governance into products their customers already pay for.

MCP gateway startups are proliferating. MintMCP, Bifrost, TrueFoundry, Obot, Composio – all attacking “connect agents to tools securely” with venture backing. Obot provides a complete open-source MCP platform including gateway, catalog, and agent orchestration.

Agent builder platforms overlap. StackAI, Dust, n8n, LangFlow, Dify, and OpenAI AgentKit offer visual workflows for multi-agent coordination. LangSmith and LangGraph cover deployment, observability, and human-in-the-loop governance.

Consulting firms are competing for advisory dollars. Accenture, Deloitte, and PwC are building MCP and A2A into their enterprise AI practices, targeting the same mid-market companies.

Large vendors are moving down into governance. Open-source frameworks are moving up into production tooling. WorkingAgents is in the middle, with better architecture than most but less distribution than all.

Verdict: The market is not empty. WorkingAgents’ governance-by-design approach (preventing unauthorized actions before they happen, not detecting them after) is genuinely differentiated. But differentiation doesn’t survive a bundling war. The question is whether WorkingAgents can land enough customers before the platforms absorb the governance feature set.

The Business Case: Challenging

Strengths

The governance wedge is sharp. “Over-privileged AI had a 76% incident rate versus 17% for least-privilege” is a stat that opens doors. Security teams – the biggest blocker of AI adoption – are the natural buyer.

The “start with Connector, graduate to Orchestrator” funnel mirrors how every successful enterprise platform has landed and expanded. Low-commitment entry, high-value expansion.

Self-hosted deployment eliminates the security review that kills most AI vendor evaluations. This is a real go-to-market advantage in regulated industries.

The founder has built the product. 86+ MCP tools, working access control, audit trails, multi-provider LLM routing, WebSocket real-time, knowledge base – this is not vaporware. Solo founders who ship working systems are rare.

Weaknesses

Solo founder. One person building, selling, deploying, and supporting an enterprise product. The most likely failure mode is not insufficient intelligence – it’s dilution of focus. Every week risks becoming a blur of custom integration requests, random partner calls, unfocused demos, and fragmented coding. The company becomes a smart consulting operation instead of a product company.

Elixir is a hiring constraint. The talent pool is small. Enterprise clients may be uncomfortable with a stack their teams can’t maintain. Partnerships (like WyeWorks) can mitigate this, but it’s a real friction point.

No customers yet. The product works. Nobody is paying for it. The gap between “working product” and “paying customer” is where most startups die. Enterprise sales cycles in this space are 6-18 months.

No funding. Competing against venture-backed startups (Grip has $13M, Composio has raised, Obot is open-source with backing) and platform giants (Microsoft, Google, Amazon, Salesforce) without funding limits marketing, hiring, and sales capacity.

The Elixir ecosystem is small. Fewer contributors, fewer libraries, fewer developers who can evaluate or contribute to the platform. Python and JavaScript projects attract 10-100x more community engagement.

Revenue Model

The internal customer lifecycle playbook outlines three tiers:

Tier Target ARR
Enterprise $50K-$250K+ budget From $75K/year
Pro Plus $25K-$100K budget $42K/year ($3,499/month)
Pro $5K-$25K budget $6K/year ($499/month)

This is a credible pricing structure for the governance category, but it requires enterprise sales capability that a solo founder can’t sustain long-term.

Probability Assessment

Scenario 1: Standalone Product Company (15-25%)

Build the product, hire a sales team, close enterprise deals, raise venture funding, scale. This is the venture path. It requires $2-5M in seed funding, 18-24 months of sales cycle development, and competing head-to-head with funded startups and platform giants.

Why it’s hard: Solo founder, no customers, no funding, intense competition. Enterprise buyers prefer established vendors. The sales cycle is longer than most founders can self-fund.

Scenario 2: Consulting + Product Hybrid (35-45%)

Use the platform as the technical backbone for an AI consulting practice. Deploy WorkingAgents for clients as part of integration engagements. Revenue comes from consulting fees plus platform licensing.

Why it works: James is a strong engineer who can deliver. The platform is real. Consulting generates revenue immediately while building the customer base. Partnerships with firms like WyeWorks provide Elixir development capacity and client access.

Why it’s limited: Consulting doesn’t scale. Revenue is linear with hours worked. The company risks becoming “a very smart consulting operation instead of a product company with strategic value.”

Scenario 3: Acquisition or Partnership (25-35%)

Build enough product and market presence to be acquired by a platform player (Grip, Swapcard, a hyperscaler) or to become the governance layer embedded in a partner’s stack (xpander, ClearML, Baseten).

Why it works: The architecture is genuinely differentiated. Capability-based access control compiled into modules at build time is hard to replicate. The Elixir/OTP foundation provides reliability guarantees that Python-based competitors struggle to match. A well-funded acquirer gets a working governance layer instead of building one.

Why it’s uncertain: Acquisition requires visibility. A solo founder in Vietnam building in Elixir is invisible to most acquirers unless they ship something that gets attention.

Scenario 4: Open Source + Commercial (20-30%)

Open-source the governance layer to build community and adoption. Monetize through managed hosting, enterprise support, and premium features.

Why it works: MCP and A2A are open protocols. An open-source governance layer on top of open protocols is a natural fit. It attracts contributors, builds credibility, and creates distribution.

Why it’s risky: Open-source users are not paying customers. AWS has a documented pattern of taking successful open-source projects and offering managed versions. A solo founder maintaining open-source while closing enterprise deals is a resource allocation conflict.

What Would Increase the Odds

Land one design partner. One real company using WorkingAgents in production, with measurable results, changes everything. “Over-privileged AI incidents dropped from 76% to 17% after deploying WorkingAgents” is a stat that sells itself.

Pick one vertical. Healthcare, fintech, or legal. Don’t try to be everything to everyone. HIPAA, SOC 2, or EU AI Act compliance makes governance mandatory, not optional.

Partner for distribution. WyeWorks for Elixir engineering capacity. xpander or ClearML for agent runtime distribution. A systems integrator for enterprise access. The product is ready – distribution is the bottleneck.

Find a co-founder. Enterprise sales, partnerships, and fundraising require a business-focused co-founder. The strongest technical founder in the world can’t close enterprise deals while also shipping features.

Ship the Connector first. The MCP Gateway is a simpler product with a faster sales cycle. “Give your AI the same keycard you carry” is a one-sentence pitch. Land Connector customers, prove governance value, then upsell to the Orchestrator.

Bottom Line

WorkingAgents is a technically strong product addressing a real and growing market need. The governance gap in enterprise AI is documented, measurable, and getting worse as agent adoption accelerates. The architecture is genuinely differentiated – capability-based access control, Elixir/OTP reliability, self-hosted zero-egress deployment.

The risks are execution, not vision. Solo founder bandwidth, enterprise sales cycles, intense competition from funded startups and platform giants, and the Elixir talent constraint all work against a standalone product company path.

The most realistic path to success is the consulting-plus-product hybrid: deploy WorkingAgents for clients, generate consulting revenue, build case studies, then leverage those into either venture funding or a strategic acquisition.

The market window is open. The EU AI Act deadline (August 2026) creates urgency. Enterprise security teams are blocking AI adoption because governance tooling doesn’t exist. WorkingAgents exists. The question is whether one person can move fast enough to capture the opportunity before the platforms close the gap.

Overall probability of meaningful success (revenue-generating business within 18 months): 35-45%.

That’s not a bad bet. Most startups have worse odds with more resources. The product is real, the market is real, and the timing is right. What’s needed now is not more building – it’s distribution, partnerships, and one customer who proves the model works.

Sources (from WorkingAgents knowledge base):