TL;DR
- AURA combines an Apache 2.0 execution harness with active telemetry, while Resolve AI uses a proprietary knowledge graph built separately for each customer.
- AURA gives you stronger architectural control. You can inspect its code and reasoning, connect your preferred models, and run it on customer-owned infrastructure.
- AURA charges no license fee and works as a first responder that investigates incidents, proposes approved remediation, and preserves reusable investigation memory.
- Resolve AI offers a more turnkey enterprise SaaS experience, with reported adoption by Coinbase, DoorDash, MongoDB, MSCI, Salesforce, and Zscaler.
- Cloud-native teams that value inspectability and extensibility should favor AURA. Large regulated enterprises committed to established ITSM workflows may prefer Resolve AI’s enterprise polish and proof points.
What each product actually is
Mezmo built AURA as an Apache 2.0 licensed open-source SRE agent harness made for autonomous execution. AURA runs on customer-owned infrastructure, where it investigates incidents and remediates them through human approval gates. Its operational memory helps engineers harden production systems against repeat failures.
Resolve AI built a proprietary enterprise SaaS agent that connects to existing observability tools and creates a bespoke knowledge graph for each customer. Customers cannot inspect, tune, or extract that internal graph. Resolve AI has attracted substantial funding and reports customers including Coinbase, DoorDash, MongoDB, MSCI, Salesforce, and Zscaler.
At a glance: AURA vs. Resolve AI
AURA favors full ownership and control of configuration, unbounded extensibility and integration, and end-to-end inspectability and transparency. Resolve AI favors a managed enterprise experience.
How we compared them
We compared AURA and Resolve AI across architecture, deployment, model flexibility, telemetry handling, pricing, and compounding memory. These dimensions shape the control, cost, and operational limits buyers face after implementation.
Mezmo built AURA, so its technical details rely partly on Mezmo documentation. Independent information about Resolve AI’s pricing and underlying model stack remains limited, and we treat vendor claims accordingly.
Architecture: open execution harness vs. closed knowledge graph
AURA gives engineers direct access to the execution layer that investigates and acts on production incidents. Mezmo publishes AURA under the Apache 2.0 license, so you can read its code, review workflow configurations, and audit how the agent reached a conclusion. You can also inspect proposed actions before approving them. Version-controlled workflows make changes visible to reviewers instead of hiding operational logic behind a vendor interface.
AURA works through an Understand, Act, and Improve loop. The agent gathers evidence and tests possible causes during Understand. Act routes remediation through human approval gates, while Improve turns useful findings into memory for later investigations. Mezmo can curate task-scoped telemetry before the open-source or platform-hosted agents reason over it, which gives AURA focused evidence instead of a raw stream of logs and metrics.
Resolve AI keeps its incident reasoning inside a proprietary, customer-specific knowledge graph. Resolve builds that graph from each customer’s connected systems and uses it to test hypotheses about an incident. The product reads telemetry through existing observability and incident-management integrations rather than controlling the underlying pipeline. Independent product analyses describe Resolve as an integration-based overlay without its own observability backend.
Resolve customers can review the findings and outputs exposed through the product, but they cannot inspect the graph’s internal construction, modify its reasoning code, or extract the full reasoning layer for independent operation. Some enterprises may accept those limits in exchange for a managed product. Engineers who need to audit production actions, extend workflows, or retain control over the execution layer gain more architectural freedom with AURA.
Deployment: customer-owned infrastructure vs. proprietary SaaS
AURA runs inside infrastructure you control. Its Apache 2.0 licensed execution harness supports centralized deployment as a container, a Kubernetes workload through Helm, or a Rust library embedded in another application. Centralized deployments make it easy to embed and enrich existing incident response workflows with automated AURA-based investigations. You can also run AURA as a local assistant through a portable CLI distributed as a self-contained binary. The CLI can also connect to remote, centrally managed AURA instances for ad-hoc sessions.
Customer-owned deployment gives you control over network boundaries, data location, access policies, and software updates. AURA can operate in an air-gapped environment when its model provider and MCP servers remain locally reachable. Self-hosting can support strict compliance requirements, but you remain responsible for configuring and documenting the required controls.
Resolve AI follows a proprietary, contact-sales SaaS model. The reviewed public material does not document an equivalent self-hosted or air-gapped deployment path. Resolve publicly states support for SOC 2 Type II, GDPR, and HIPAA, which may simplify vendor review for enterprises that prefer a managed service and accept external processing within approved boundaries.
The deployment choice determines who carries operational responsibility. AURA gives you more control and audit access, while requiring you to operate the software. Resolve AI manages the service for you, but its enterprise sales process and closed deployment model provide less direct control over the runtime.
Model flexibility: bring-your-own-model vs. custom-trained per environment
AURA gives you direct control over the model that handles production data and executes workflows. Its public documentation lists support for OpenAI, Anthropic, AWS Bedrock, Google Gemini, Ollama, and OpenRouter, and any OpenAI compatible completions endpoint works. The provider and model are defined in a TOML configuration file, so switching providers does not require changes to application code. Ollama support also lets you run local models when data residency or network isolation rules prevent external API calls. Multiple providers and models can be assigned to individual workers, which lets you route cheaper models to less complex tasks and reserve stronger models for harder reasoning work.
Resolve AI takes a managed, customer-specific approach. The public information reviewed for this comparison does not disclose its underlying models, training methods, or whether customers can select a provider. Resolve AI may handle those choices during an enterprise deployment, but buyers cannot verify model portability or bring-your-own-model support from public documentation.
AURA offers the clearer choice when you need model choice, local inference, or protection against provider lock-in. Resolve AI suits buyers who prefer the vendor to manage the model layer. Prospective Resolve AI customers should ask which providers process their data, whether models can run in their environment, and what happens if they later need to change providers.
Telemetry approach: owning your worker orchestration vs. adapting to a rigid workflow
Resolve AI gathers incident context by querying connected observability and response tools. Independent testing describes Resolve AI as an overlay that relies on providers such as Datadog, Grafana, and PagerDuty. The same testing found that response speed depends on those tools’ APIs, while accuracy depends on the data they expose. Slow queries, omitted fields, and inconsistent retention can therefore limit an investigation.
AURA workers can be defined to use tools from any external connection over MCP or A2A protocols, and external vector stores are supported for retrieval-augmented generation. External connections are not limited to a predefined list of observability services. For example, workers can search Slack history, review Jira tickets, and inspect recent source control changes. This gives you an almost unlimited degree of flexibility in choosing investigative sources that mirror what a human operations analyst would use, not just telemetry. AURA then curates task-scoped context for each investigation instead of filling a context window with unfiltered telemetry.
Mezmo customers can also benefit from telemetry that is pre-distilled for use with agentic SRE tools like AURA. Anomalous logs, metrics, and traces can be automatically distilled into an AI-ready investigation package, available through the Mezmo MCP. This packaging can also be triggered on demand through an external trigger like a PagerDuty incident, which creates a meaningful efficiency gain on MTTA and MTTR.
Resolve AI fits environments that already centralize reliable data in supported tools and want an agent layered over them. AURA fits environments where you want to control how telemetry and collaborative or code context becomes evidence, particularly when several models or operational tools need to use the same context. Mezmo customers get the added benefit of high-speed root cause analysis backed by token-efficient telemetry context.
Pricing: free and open source vs. enterprise SaaS
AURA uses the Apache 2.0 license and charges no software license fee. You can evaluate the code, run a pilot, or scale to production use without engaging a sales process or waiting on budget approval. You still pay for infrastructure, models, integration work, and the engineers who operate it.
Resolve AI sells proprietary SaaS through a custom enterprise quote. Resolve does not publish standard prices, so buyers need to contact sales before they can compare total costs or begin a formal purchase. Its named large-enterprise customers indicate a sales motion built for organizations with established procurement and vendor review processes.
The practical difference appears before deployment. AURA lowers the budget barrier to initial testing but asks you to own the implementation. Resolve places sales and procurement earlier in the evaluation, while its managed service can reduce the internal work required to get an enterprise deployment running. Buyers should still budget time for ramp-up after signing, since expanding AI use safely into production incident response is rarely instant with any vendor.
The compounding question: substrate vs. bespoke agent
AURA compounds through reusable investigation memory and a shared execution substrate. After each run, AURA compresses the evidence, findings, and decisions into a summary that later investigations can reuse. The agent spends fewer tokens rebuilding known context, which can reduce the cost of investigating recurring failures over time.
AURA also lets improvements travel between deployments. Engineers can review and contribute to its open-source codebase, while organizations can iterate on version-controlled configurations and workflow patterns and share them with the wider community. Each organization keeps its operational data private, but improvements to the harness and shared best practices remain an ongoing, always-improving resource for others to inspect and adopt.
Resolve AI builds a bespoke knowledge graph for each customer. That graph can become more useful as it learns one environment, but customers cannot inspect, tune, or extract the underlying graph. Knowledge accumulated inside one customer deployment does not become a portable asset that another deployment can adopt.
Buyers should evaluate how each approach holds up beyond the current demo. Resolve AI concentrates value in a vendor-managed agent rebuilt around each environment. AURA concentrates value in an open substrate plus customer-owned memory, so you can retain prior investigation knowledge while adopting improvements to the shared harness.
Resolve AI's enterprise traction, credited honestly
Resolve AI has credible momentum among large enterprises. In its Series A announcement, the company reported a $125 million round at a $1 billion valuation, bringing total funding above $150 million. The same announcement named Coinbase, DoorDash, MongoDB, MSCI, Salesforce, and Zscaler as customers. The supplied sources do not substantiate newer figures of $190 million raised or a $1.5 billion valuation.
Resolve AI also reports measurable customer outcomes. Coinbase reduced its time to root cause by 72 percent, while Zscaler reduced the number of engineers required per incident by 30 percent. These figures come from Resolve AI's founders, employees, and investors. No independent audit or customer-published study appears in the supplied research, so buyers should treat them as vendor-reported results.
Resolve AI's managed SaaS model can fit large enterprises that already route incidents through formal IT service management workflows and established observability tools. Those buyers may prefer a vendor-operated agent over owning an open-source execution harness. However, the supplied sources do not verify specific compliance certifications or ITSM integrations. Regulated buyers should confirm data residency, audit controls, deployment options, and integration depth during procurement.
Who should choose which
Choose AURA if you run cloud-native infrastructure and want direct control over the agent that can act in production. Its Apache 2.0 license lets you inspect the code, deploy it on your infrastructure, and audit its reasoning. AURA also lets you choose models and orchestration tools, while human approval gates govern remediation.
Choose Resolve AI if your enterprise already centers incident operations on established ITSM workflows and observability vendors. Resolve AI packages investigation inside a proprietary SaaS product, which can reduce the engineering work required to assemble and maintain an agent harness. Buyers that prioritize vendor-managed delivery and enterprise references over inspectability will likely prefer this approach.
Regulated enterprises should base the decision on their specific controls rather than company size alone. AURA gives you stronger control over deployment, code inspection, and model choice. Resolve AI may fit an established procurement model better, but buyers should verify its current certifications and data-handling terms because the available research does not establish those details.
FAQ
Is AURA a replacement for Resolve AI?
A replacement must cover the same operational role without requiring the same architecture. AURA can replace Resolve AI when you prefer an open, self-hosted first responder over proprietary SaaS. You retain control over models, workflows, reasoning, and production actions.
Can AURA and Resolve AI be used together?
Coexistence lets two agents work against the same production stack. AURA can connect to your stack while Resolve AI reads existing observability tools, although neither vendor documents a native connection between them. You can evaluate both on bounded incident types before standardizing.
Which fits regulated enterprises better?
Regulated fit depends on your compliance controls and operating model. Resolve AI suits enterprises that prioritize managed SaaS and established ITSM workflows, while AURA suits those that require code inspection and customer-owned deployment. Your security and audit requirements should determine the choice.
How do licensing and cost differ?
Licensing determines what you may run, modify, and redistribute. AURA carries an Apache 2.0 license with no license fee, while Resolve AI uses custom-quote enterprise SaaS pricing. AURA lowers the initial budget barrier, but you still pay for infrastructure, models, and operations.
