The facts, with their sources

What it does

Writes code: PartlyCode generation is named as one task the agents LeewayHertz builds can execute, so authoring code is a capability of delivered solutions rather than a product you buy; the firm itself also writes the solution code under an engagement.we build intelligent AI agents capable of executing a multitude of tasks, including analysis, research, code generation, audits, and reviewsleewayhertz.com
Writes tests: PartlySoftware testing agents are listed among the IT-function agents LeewayHertz builds, and its own delivery includes unit, integration and performance testing, but no page claims automated test authoring as a standing capability.leewayhertz.com/ai-agent-development-co…
Reviews code: PartlyThe homepage states the agents it builds can perform audits and reviews alongside code generation; the vendor does not describe a diff-review workflow on a pull request, so this holds for custom-built agents only.capable of executing a multitude of tasks, including analysis, research, code generation, audits, and reviewsleewayhertz.com
Ships it: YesLeewayHertz runs deployment itself, including CI/CD pipelines, containerised and controlled release processes, so engagements go past design into production operation rather than stopping at a design document.We implement continuous integration and delivery (CI/CD) pipelines to automate the deployment process. This ensures that your AI solutions are…leewayhertz.com/ai-deployment-services-…
Done-for-you service: YesThis is a services business: LeewayHertz people run the full agent lifecycle for the client, from use-case analysis through development to AgentOps, rather than selling software the buyer operates alone.We support the complete agent lifecycle from use-case analysis and technical design to development, governed deployment, and AgentOps.leewayhertz.com/ai-agent-development-co…

How much it does unattended

Runs autonomously: PartlyThe vendor claims solutions that automate multi-step workflows with minimal human intervention, but the same sentence keeps defined human approval and oversight in the path, so autonomy is bounded by design per engagement.We build agentic AI solutions that reason, orchestrate tasks, use enterprise data and tools, and automate complex, multi-step workflows with minimal…leewayhertz.com/ai-consulting-services-…
Multi-agent: YesMulti-agent system development is a named service, with sequential, hierarchical, supervisory or event-driven coordination patterns chosen per workflow and explicit handoffs between specialised agents.Yes. We develop multi-agent systems for workflows that require specialized responsibilities, parallel activities, independent validation, or…leewayhertz.com/ai-agent-development-co…
Agent permissions: PartlyHow much an agent may do is set per build: identities, permitted tools, autonomy boundaries, human approvals, escalation paths and stop controls are configured, so there is no fixed product-wide permission model.Agents are configured with defined identities and permissions, access to approved tools and data, autonomy boundaries, human-approval requirements,…leewayhertz.com/ai-agent-development-co…

Which models it runs on

Claude: YesClaude is named among the foundation models LeewayHertz selects and fine-tunes for domain-specific work, and Anthropic appears in its agent-development model stack, so client builds can run on it.we select and fine-tune a foundation model like GPT, Llama, PaLM, Claude, Gemini, Mixtral or Mistral and integrate it with your proprietary dataleewayhertz.com/ai-consulting-services-…
GPT: YesGPT is the first model named in the vendor's own foundation-model sentence, and OpenAI sits in the models-and-model-services row of its agent tech stack, so GPT-backed builds are standard.we select and fine-tune a foundation model like GPT, Llama, PaLM, Claude, Gemini, Mixtral or Mistral and integrate it with your proprietary dataleewayhertz.com/ai-consulting-services-…
Open models: YesOpen-weight families Llama, Mixtral and Mistral appear in the same fine-tuning sentence as the closed models, so engagements can be delivered on open models rather than hosted APIs only.we select and fine-tune a foundation model like GPT, Llama, PaLM, Claude, Gemini, Mixtral or Mistral and integrate it with your proprietary dataleewayhertz.com/ai-consulting-services-…
Model choice: YesThe firm states a technology-agnostic approach where the stack is picked for the use case, enterprise standards and operating cost, and its architecture isolates a model layer with routing and fallback options.Our technology-agnostic approach ensures the right technology stack is chosen for the use case, enterprise standards, deployment requirements, and…leewayhertz.com/ai-agent-development-co…

Where you use it

In a browser: PartlyDelivered solutions can expose chat, voice, embedded application interfaces, work queues and review screens, so a web surface is a design choice in the experience layer rather than a hosted product you log into.Experience Layer: for chat, voice, embedded application interfaces, work queues, notifications, review screens, and other user interaction channels.leewayhertz.com/ai-agent-development-co…

Whose machine it runs on

Self-hosted: YesOn-premises deployment is an explicitly offered option for buyers needing full control of infrastructure and data, alongside hybrid, multi-cloud and container-based deployment.For businesses that require full control over their AI infrastructure and data, we provide on-premises deployment solutions.leewayhertz.com/ai-deployment-services-…

What it costs to run

How it meters: PartlyBilling is by engagement model, dedicated team, team extension or project-based, with no seat, credit or token unit published anywhere on the site, so the commercial unit is negotiated per contract.Dedicated Development Team ... Team Extension ... Project-based Modelleewayhertz.com/ai-agent-development-co…
Free tier: NoThere is no self-serve way in at all: every call to action routes to a contact form and a scheduled call under NDA before any work or quote, and no trial or sandbox is offered.Once you let us know your requirement, our technical expert will schedule a call and discuss your idea in detail post sign of an NDA.leewayhertz.com/ai-agent-development-co…
MCP server: PartlyMCP appears only as a protocol LeewayHertz implements inside client builds for tool and context access; no public MCP server or connector of its own is documented, so another agent cannot query the firm.Model Context Protocol (MCP) for standardized access to tools and contextleewayhertz.com/ai-agent-development-co…
Bring your own key: YesSolutions are packaged and deployed into the client's own cloud, tenant, region and security perimeter, with enterprise-controlled model endpoints in the model layer, so the buyer keeps the inference account.We package and deploy agentic solutions within the organization’s preferred cloud, tenant, region, and security perimeter.leewayhertz.com/ai-agent-development-co…

Buying it for a team

A company can buy it: YesThe entire offer is sold to organisations, enterprises and startups, through proposals and engagement models; there is no individual self-serve purchase path on the site.We help enterprises build and deploy AI agents and multi-agent systems that automate complex workflowsleewayhertz.com/ai-agent-development-co…
Admin controls: YesConstraining the agent is a core part of the build: least-privilege access, policy checks, confidence thresholds, guardrails, approvals and stop controls limit which systems and actions are available to it.Least-privilege access limits the systems, information, environments, and actions available to the agent. Additional controls can include policy…leewayhertz.com/ai-agent-development-co…
Audit log: YesRuntime monitoring, audit trails and alerts are stated as delivered controls, and traceability retains inputs, sources, tool activity and approvals needed to review how an outcome was produced.Runtime monitoring, audit trails, alerts, incident procedures, and stop controls provide visibility into agent activity and enable interventionleewayhertz.com/ai-agent-development-co…
Single sign-on: PartlyEnterprise identity providers and OAuth 2.1 are listed in the security and protocol stack LeewayHertz builds with, but neither SAML nor OIDC is named, and sign-on depends on the client's own estate.leewayhertz.com/ai-agent-development-co…

What happens to your code

Data residency: PartlyRegion is chosen at deployment inside the client's own perimeter, which answers residency per engagement, while the firm's own website privacy policy warns personal data may be transferred to other jurisdictions.We package and deploy agentic solutions within the organization’s preferred cloud, tenant, region, and security perimeter.leewayhertz.com/ai-agent-development-co…