conw
Conw is a locally-served AI chatbot built on the Conway-Retrain 12B model and served via MLX inference on your own hardware. The system implements a guarded learning loop where team members correct replies, rate them, and explain new vocabulary; unsafe or low-quality responses are filtered automatically before verified examples update the live model. All learning remains private to your account. The model is accessible via an OpenAI-compatible API, allowing Developers to integrate Conw into custom tools and workflows without external API calls or data routing to third-party providers.
conw is an AI chatbot, priced at £15 a month on the Pro plan, integrating with OpenAI SDK and Hugging Face. InnovaAI rates it 4.6 of 10 for agency adoption, best for Developer, Operations Manager and Project Manager roles.
Agency Audit
Conw is a locally-hosted AI chatbot built on the Conway-Retrain 12B model that learns from your team's corrections and feedback without routing data to external APIs. Developers and operations teams benefit most, since Conw integrates via OpenAI-compatible API for custom workflows and keeps all learning private to your account. The guarded learning loop, where unsafe or low-quality responses are filtered before updating the live model, makes it suitable for agencies building proprietary AI assistants or handling sensitive client work. Adoption pays off if your team runs custom AI integrations or prioritizes data privacy over out-of-the-box convenience.
5recommended
60/mo
$4,480/mo
Moderate
Illustrative scenario. Not a guarantee. Net capacity is the value of reclaimed time at $75/hr, less the lowest verified paid base plan (flat plan cost is shared). Hours saved come from the service estimate; implementation, taxes, and unprovided usage charges are excluded.
- Developer handling custom AI tool integration
- Operations Manager handling proprietary model training and feedback
- Project Manager handling client-specific vocabulary management
- Your team expects a plug-and-play AI assistant without active feedback loops; Conw's value compounds only if you consistently correct replies and explain new vocabulary.
- Your agency has no in-house Developer capacity to integrate the OpenAI-compatible API or manage locally-served inference on your hardware.
- You need real-time transcription or meeting automation; Conw is a chatbot, not a meeting-capture tool, and does not include audio ingestion or call recording.
Internal Adoption Path
$19.98/mo
$19.98/mo flat plan
60 hr/mo
5 seats × 12 hr each
$4,500/mo
modeled at $75/hr labor rate
$4,480/mo
value − subscription cost
In this model, 5 seats reclaim 60 hours of team time each month. Valued at $75/hr that is $4,500/mo, and after the $19.98/mo subscription it leaves $4,480/mo of capacity for billable client work.
Illustrative scenario. Not a guarantee. Uses the lowest verified paid base plan. Implementation, taxes, and unprovided usage charges are excluded.
Platform Features
Core capabilities of conw
Guarded learning loop with safety filtering
Team members correct replies and rate them; unsafe or low-quality responses are automatically filtered out before verified examples update the live model. Developers and Operations teams use this to ensure proprietary AI assistants stay accurate and safe without manual review overhead.
OpenAI-compatible API for custom integrations
Developers integrate Conw into internal tools and client workflows without external API calls or data routing to third-party providers. This eliminates the need to choose between convenience and data privacy when building custom AI features.
Private vocabulary memory per account
Team members explain new words and terminology; the model learns agency-specific language without sharing adapter candidates between accounts. Account Executives and Project Managers benefit by getting responses that reflect your client base and industry jargon.
Locally-served inference on your hardware
The Conway-Retrain 12B model runs via MLX on your own infrastructure, keeping all chat history and learning data off third-party servers. Operations teams reduce compliance friction and latency for sensitive client work.
Learning dashboard with pass/failure visibility
Team members see which corrections were promoted to the live model and which were filtered out, making the feedback loop transparent. Project Managers use this to track model improvement over time and identify gaps in training data.
Chat history, pinning, and search
Team members organize and retrieve past conversations without relying on external note-taking tools. Account Executives and Project Managers save time on context-switching by keeping all AI-assisted work in one searchable interface.
What Makes conw Different
Unique advantages vs similar tools in this niche
Transparent learning loop with visible pass/fail states
vs Opaque AI models that don't show how they learnConw shows whether a lesson passed or failed, and keeps useful memory even when weight updates are rejected.
Locally served model for privacy and efficiency
vs Cloud-dependent AI assistants that forward data to external APIsConw runs on a single 16GB iMac, not a GPU cluster, and does not forward questions to an external answer API.
OpenAI-compatible API with pay-as-you-go pricing
vs Proprietary APIs that require custom SDKsDrop-in OpenAI-compatible endpoints allow swapping base_url and going.
Value Equation
Outcome-likelihood-time-effort assessment for conw
Limited agency channel
conw scored below the agency-resellability threshold (agency_fit_score < 50). The Value Equation projects agency-side outcomes, which don't apply to tools without a clear resell pathway.
Contact conwPricing
conw platform cost to your agency
Starts at £15/mo (Pro), scales to £30/mo (Max)
Free
- 25K tokens per week
- The full Conway-Retrain 12B model
- Private vocabulary memory that works immediately
- Verified per-user adapter candidates, never shared between accounts
Pro
- 500K tokens per week
- 20× the Free allowance
- The full Conway-Retrain 12B model
- Private vocabulary memory that works immediately
Max
- 2M tokens per week
- 4× Pro · 80× Free allowance
- The full Conway-Retrain 12B model
- Private vocabulary memory that works immediately
No verified white-label program for conw: client-facing delivery runs under the platform's native branding.
Prices as published by the vendor in GBP · your regional price may differ
Market Intelligence
Offer + scale economics for conw
Limited agency channel
conw scored below the agency-resellability threshold (agency_fit_score < 50). It's a useful tool but not designed for white-labeled or retainer-based reselling, so we don't publish productized offer economics for it.
Contact conwInvestment Decision Framework
Strategic vetting analysis for conw
Situational Fit
Fit depends on your client mix
Buy If
4Your Developers spend 3+ hours per week integrating third-party AI APIs into client tools and want to avoid external data routing by using the OpenAI-compatible API instead.
Your Operations team manages sensitive client data and needs an AI assistant that keeps all learning private to your account rather than feeding corrections into a shared cloud model.
Your Project Managers or Account Executives regularly need custom AI workflows tailored to your agency's vocabulary and terminology, and you're willing to invest time in the learning loop to train the model.
Your team builds AI-powered features for clients and wants to host the inference layer on your own hardware to reduce latency and third-party dependencies.
Skip If
4Your team expects a plug-and-play AI assistant without active feedback loops; Conw's value compounds only if you consistently correct replies and explain new vocabulary.
Your agency has no in-house Developer capacity to integrate the OpenAI-compatible API or manage locally-served inference on your hardware.
You need real-time transcription or meeting automation; Conw is a chatbot, not a meeting-capture tool, and does not include audio ingestion or call recording.
Your budget cannot accommodate the infrastructure cost of running MLX inference on dedicated hardware, or your IT team cannot support a locally-hosted model.
Bottom Line
Conw is a locally-hosted AI chatbot built on the Conway-Retrain 12B model that learns from your team's corrections and feedback without routing data to external APIs. Developers and operations teams benefit most, since Conw integrates via OpenAI-compatible API for custom workflows and keeps all learning private to your account. The guarded learning loop, where unsafe or low-quality responses are filtered before updating the live model, makes it suitable for agencies building proprietary AI assistants or handling sensitive client work. Adoption pays off if your team runs custom AI integrations or prioritizes data privacy over out-of-the-box convenience.
Reality Check
Conw requires your team to actively correct and rate responses to improve the model, which means adoption friction if your workflow doesn't naturally include feedback loops. The locally-served model runs on your own hardware, adding infrastructure responsibility compared to cloud-only alternatives.
Low effort: self-service setup with guided onboarding
Academy for conw
Work through it in order: the course for this service first, then the modules behind it.
Course for this service
Conw Agency Implementation, Private AI Chatbots for Client Workflows
Learn how to deploy Conw's locally-hosted 12B model as a white-label chatbot service for clients. This course covers setting up private vocabulary memory, building the guarded learning loop with your team, integrating via OpenAI-compatible APIs into custom client tools, and packaging these capabilities into retainer-based offerings without external API dependencies.
Open the courseNo Academy modules are published for this service yet. Browse the full Academy
Why this category matters
The commercial case before the tooling.
Core concepts
The mental model you need to price and scope the work.
- Escalation Debt RatioConcept
Escalation Debt Ratio is the share of chatbot conversations that must reach a human before resolution, measured against the share the client was promised would resolve alone. Agencies sell AI chatbots on deflection rates, but the retainer survives on the escalation path: who answers at 2am, how fast, and whether the transcript carries context. A bot trained on client documents can close routine questions, yet pricing, refunds, and account changes usually need a person. Track the ratio monthly. When it climbs, the client sees a queue, not savings. Chatling and FastBots both ship human handover as a first-class feature, which tells you the vendors expect escalation to be normal, not exceptional. The framework matters because escalation debt compounds quietly: every unresolved thread becomes a support ticket, a churn signal, and a reason the client questions the retainer. Budget the human layer before you quote the automation layer.
- Handoff Fidelity ThresholdConcept
Handoff Fidelity Threshold is the point at which a chatbot's automated resolution stops being cheaper than a clean transfer to a human. Below the threshold, every extra automated turn saves the agency money. Above it, each additional bot turn adds rework, apology credits, and churn risk that the retainer has to absorb. The framework forces agencies to define the threshold per client before deployment, not after the first complaint. A support bot trained on uploaded documents and website content can resolve repetitive questions well, but the moment a query touches billing disputes, refunds, or account-specific history, the cost curve inverts. The practical test: measure containment rate and escalation satisfaction separately, because a 70 percent containment rate with angry escalations is worse economics than 50 percent containment with clean transfers. Agencies that price managed chatbot retainers on containment alone are selling a metric that hides the expensive half of the conversation.
- Automation Coverage CeilingConcept
Automation Coverage Ceiling is the share of inbound conversations a chatbot can resolve without a human, and it is almost never 100%. Agencies that sell full automation promise a number the client's own question mix will not support. The ceiling is set by three things: how repetitive the inquiry set is, how much judgment a correct answer requires, and how fast the knowledge base decays. A support desk answering shipping-status questions can sit near 80% deflection; a services firm fielding scoping calls may top out near 40%. The framework matters because the gap between the ceiling and 100% is exactly where the human escalation path, and the agency's ongoing optimization retainer, live. Chatling and FastBots both ship human handover as a first-class feature for this reason. Price the managed tier against the residual, not the total volume.
Decision and risk
How to judge the fit, and the ways it goes wrong.
- When Chatbot Scope Outruns Escalation Design, Cap the Automation Before You Sell the RetainerEvaluation Rule
Cap the automation scope at the intents you can escalate cleanly, and price the retainer around ongoing optimization rather than full deflection.
- When Clients Buy Full Automation, Price the Escalation Path FirstEvaluation Rule
Price the escalation path as a named retainer line item before you quote the bot build, and refuse the deal if the client will not staff it.
- Managed Chatbot Retainer vs One-Off Bot BuildDecision Framework
IF a client's inquiry volume is steady, their knowledge base changes monthly, and they lack an internal owner for escalation rules, THEN sell a managed chatbot retainer with defined optimization hours rather than a one-time build. IF the client has a single campaign window, a frozen FAQ set, and an in-house ops lead who will own the bot after handover, THEN a fixed-scope build is the cleaner commercial fit.
- The Full-Automation Trap: Why AI Chatbot Retainers Stall When Escalation Paths Are Never BuiltFailure Pattern
- The Pilot Purgatory Trap: Why AI Chatbot Deployments Stall Before They Reach a RetainerFailure Pattern
Delivery system
Blueprints and procedures for running it as a service.
- Managed AI Chatbot Retainer Launch (10-21 days)Implementation Blueprint
A productized engagement that takes a client from raw FAQ and policy documents to a live, escalation-aware chatbot with a monthly optimization retainer attached. Built for agencies that want recurring revenue instead of one-off build fees.
- Escalation Path Design (Onboarding)Operating Procedure
- Conversation Transcript Review (QA)Operating Procedure
- Knowledge Base Intake and Training Data Gate (Onboarding)Operating Procedure
13 modules selected for conw
Frequently Asked Questions
Answers about pricing, setup, implementation
Conw is a locally-served AI chatbot that learns from your team's corrections and feedback. Team members rate replies, explain new vocabulary, and filter unsafe responses; verified examples update the live model after safety checks. The model is accessible via OpenAI-compatible API, allowing Developers to integrate Conw into custom tools without routing data to external providers.
Pro plan is £15 per seat per month and includes 500K tokens per week. Max plan is £30 per seat per month and includes 2M tokens per week. A free tier is available with 25K tokens per week and the full Conway-Retrain 12B model.
Developers benefit most, since they integrate Conw via the OpenAI-compatible API into custom client tools and internal workflows without external API calls. Operations teams gain value by keeping all learning private to your account and running inference on your own hardware. Project Managers and Account Executives improve context retention by using the searchable chat history and learning dashboard to track model improvement over time.
Conservative estimate is 3 to 5 hours per week per Developer seat on API integration and data-privacy workflows, since you eliminate the need to route data through third-party AI services or manage multiple API keys. For Operations and Project Management roles, savings depend on how much time your team currently spends managing external AI tools or documenting client-specific vocabulary; payback is highest if your team spends 4+ hours per week on those tasks.
Conw exposes an OpenAI-compatible API, so Developers can integrate it into any tool or workflow that supports OpenAI SDK or Hugging Face. All learning and chat history remain on your local instance; there are no third-party integrations that route data off your hardware.
Since Conw runs on your own hardware and all learning is private to your account, you retain full control of chat history, corrections, and trained vocabulary. Cancellation does not affect data stored on your local instance.
For non-Developer roles, rollout is low-friction; team members can start using the chat interface immediately. For Developers integrating the API, rollout depends on the complexity of your custom workflows; expect 1 to 2 weeks for a basic integration and longer for multi-tool deployments.
Yes, Conw runs the Conway-Retrain 12B model via MLX inference on your own hardware. Your IT team must provision and maintain the infrastructure; this adds operational responsibility compared to cloud-only AI assistants.