Serval vs Siit (2026)

The integration catalog runs to 164 documentation pages, indexed at docs.serval.com/llms.txt, with a page per connector. Ticketing covers ServiceNow, Jira, Zendesk and Freshservice, bidirectionally, so a customer can keep existing tooling during migration. Identity covers Okta, Microsoft Entra, LDAP and Active Directory, plus SailPoint for governance and RSA SecurID for MFA. Endpoints run through Jamf and Tanium and HR through Workday.

Knowledge comes from Notion, Confluence, Google Drive and Slab, collaboration from Slack, Microsoft Teams and Google Workspace, and finance and operations connectors include Coupa, Salesforce, LaunchDarkly, Vanta and Presidio. Connectors take authenticated action rather than mirroring data.

For RSA SecurID, Serval stores the client key encrypted and attaches it automatically to every call it makes on the customer's behalf, and workflows drive a full MFA flow, starting, verifying, checking and canceling attempts on demand. For Presidio, Serval exchanges credentials for a short lived JWT and reads asset records on the customer's behalf whenever a Serval workflow runs, with the regional endpoint chosen by the workflow.

Both happen at execution rather than by background sync. Provisioning is the heaviest write path, with automated provisioning and deprovisioning via SCIM, direct API or custom workflows, and just in time access grants revoked automatically when the window ends. Access can also be revoked via API. Serval's AI knows the endpoint catalog and can help build workflows against it. Connector maturity is labeled, with RSA marked Beta in the integrations UI.

Sourcedocs.serval.com RSA SecurID and Presidio integration pages, with the integrations indexread 2026-09-12

More than 100 ready to use integration providers come as a plan entitlement. Named systems span HRIS (Workday, BambooHR, Rippling), device management (Kandji, Jamf, Microsoft Intune), identity (Okta, Microsoft Entra ID, Google, JumpCloud, Active Directory), ticketing (Jira Service Management, Zendesk) and documentation (Notion, Confluence, Google Workspace). Siit writes into systems it does not own.

Automated app access management lets a customer define who can request each app, routes approvals automatically and provisions access through the identity provider, with an audit trail on every request. Time based deprovisioning grants access for a set duration and revokes it automatically when it expires, so temporary access never lingers. Siit grants and revokes access to third party SaaS apps through the customer's identity provider, unattended and on a timer.

App discovery works the other way, filling the app list by connecting Google, Okta, Entra ID or JumpCloud. Siit's own API can add a user to an application and remove one, with the deprovisioning state shown in the response. Equipment, Application, Date, People and Request workflows each act on connected systems, and Custom Webhooks send real time data from Siit to other systems to trigger actions instantly.

Sourcesiit.io/pricing feature matrix, with developer.siit.io API referenceread 2026-09-12

A workflow can call another as a step through child workflow composition, workflow journeys can be viewed as dependency graphs, and workflows can be organized into folders, shared across teams, tagged, searched semantically and versioned with diffs. Temporal is listed among the required datastores as the workflow orchestration component. It is a durable execution engine, so a run survives process restarts and resumes, which holds long running, multi step work together when people interrupt it.

Workflows can pause for approval, and Serval also has conditional forms in Slack, criteria based assignment rules, a default triage team for ambiguous requests and ITIL aligned change management. The Automation Agent turns natural language descriptions into TypeScript, so automations build themselves from a description. Serval markets deterministic workflows rather than execution the model improvises, since the model writes the workflow and the workflow then runs the same way every time. Workflows and skills are managed as code in Git with version diffs.

The participants include the employee raising a request, three named agents with different jobs, approvers who can be delegated to or overridden by a team manager, assignment groups and external ticket subscribers. Work spans ticketing, just in time access, asset management, onboarding and offboarding, compliance workflows and campaigns, with Security, IT, HR, Finance and Operations each able to run isolated environments on the same platform.

Sourcedocs.serval.com release notes index and self-hosting prerequisitesread 2026-09-12

Work is composed across systems and object types.

A no code automation engine with approval logic and fallback routing runs five workflow families, sold as separate plan entitlements (Equipment, Application, Date, People and Request), so a single onboarding flow can create a user, grant application access through the identity provider, assign equipment from inventory and open the requests that need human hands, each acting on a different object in a different system. Routing is conditional and stateful.

It can go by availability, so nothing lands with an admin outside working hours, by round robin, to level load across an inbox, or by group assignment to a team inbox. Escalation and reassignment are built in, and the API can update assignees as long as the assignee admin belongs to the assignee inbox. Flows pause for people and resume.

Request Approvals ask a manager or stakeholder to validate, and automated app access management routes approvals before provisioning, so the flow holds mid execution and continues on a human decision. The service model follows ITIL. Incident, Change and Problem Requests are separate named types, each a process with stages: Change Requests plan, track and implement IT changes with minimal risk, and Problem Requests find the root causes of recurring issues.

Each request has distinct participants, namely the requester, the admin who resolves it, approvers who are managers or stakeholders, followers kept informed, team inboxes acting together, and on the Pro tier the AI agent that resolves requests on its own. Workspaces let several teams run their own flows on one account.

Sourcesiit.io/pricing feature matrix, with developer.siit.io API referenceread 2026-09-12

Workflows can be triggered by events and by webhooks, which let any external system act as an event source, as well as by Slack reactions, SLAs and ticket assignment, and campaigns can be triggered directly from a workflow. Child workflow composition lets one flow start another, and tickets, access requests and onboarding events all start work. Schedules are a separate mechanism, covering recurring campaigns, assignment schedules, SLAs with business hours and pause conditions, and automatic ticket bumps for unresponsive requesters, a time based trigger acting without a person.

The Help Desk Agent covers intake and resolution across Slack, Teams, email, phone and portal. Further channels include an embedded widget with a live agent and multiple email intake addresses per team, along with Microsoft Teams notifications and reply routing improvements, conditional forms in Slack, dynamically rendered forms, a Serval Inbox consolidating notifications, and white label Slack bot customization. The UI ships in nine languages with on demand ticket translation.

Sourcedocs.serval.com release notes index and serval.com/insights buyer's guideread 2026-09-12

Workflows run on triggers based on requests or people, from request created or service assigned to start date, birthday or work anniversary, which makes two event triggers and three date triggers. Workflows also come in five families named for the object that fires them, Equipment, Application, Date, People and Request. Alerts are a separate mechanism.

Custom Alerts are notifications triggered by conditions like SLA breaches or inactivity, so they fire on the absence of an event as well as its presence, and SLA Management monitors response and resolution times. Time based deprovisioning revokes access when a grant expires, a scheduled action with no person present. Custom Webhooks send real time data from Siit to other systems to trigger actions instantly, so Siit is an event source as well as a consumer.

Every plan, including the entry tier, has a Slack Bot, a Microsoft Teams Bot, an Email Channel and a Self-Service Portal where employees browse the service catalog, and Broadcasts send targeted messages on Slack or by email. Routing on arrival is conditional, with availability based routing sending work only to admins inside their working hours and round robin spreading it across an inbox.

Sourcesiit.io/pricing feature matrixread 2026-09-12

A native knowledge base, in beta, sits alongside the synced stores and is maintained in Serval, with knowledge base management, knowledge tagging within guidance, lexical and keyword search alongside semantic retrieval, cross team KB sharing, and a tag catalog with owners set for teams, resources and roles. Answers ground in help docs built in Serval or synced from Notion, Google Drive, Confluence and Slab, and reviewers can publish approved knowledge updates into Notion, Slab, Confluence, Freshservice and ServiceNow, so corrections flow back out to the systems of record.

Retrieval respects permissions at read time through knowledge source permission enforcement, so an employee cannot be answered from a document they are not entitled to see. The Help Desk Agent can also search the web, tickets, assets and access history are queryable alongside, and the Insights Agent analyzes ticket patterns. Serval's asset estate tracks hardware, software licenses, SaaS subscriptions and cloud costs in one schema the agent reasons over.

Sourcedocs.serval.com release notes indexread 2026-09-12

Alongside reading other systems, Siit keeps its own knowledge store. The API manages articles through a full lifecycle, listing, creating, retrieving, updating, publishing, unpublishing, archiving and unarchiving them, and List Categories returns top level categories and subcategories with a parent field to rebuild the hierarchy. Publish and archive states give editorial control, and the category tree keeps the store curated. Synced content joins the same corpus.

The knowledge base is a central place for articles from Siit or connected systems like Confluence and Notion, with related data, on every plan including the entry tier, and Google Workspace is named alongside. The assistant draws on Notion, Confluence and Google Workspace together with Siit's own articles. Retrieval is proactive, since AI Article Suggestions scans the knowledge base and recommends relevant articles to employees before they submit a request. People records, the app inventory and the equipment inventory also give the agent structured context about who is asking and what they hold, from the customer's own application data.

Sourcedeveloper.siit.io article endpoints and siit.io/pricing feature matrixread 2026-09-12

Per team guidance, knowledge sources and configuration persist in Serval, but none of these is a memory that learns per user over time, and no agent memory is described. Guidance, now called Skills, is a set of written procedures and runbooks the customer authors so agents follow internal practice, with a diff view for suggested changes and version history. It persists because a person wrote it down, not because the agent retained anything.

The native knowledge base and the corpora synced from Notion, Confluence, Google Drive and Slab are the knowledge layer, and workflows, skills and settings managed as code in Git are authored artifacts. The runtime stack uses PostgreSQL as the primary application database, Redis for caching and ephemeral state, and Temporal for workflow orchestration.

Temporal keeps a workflow run alive through restarts, but that is state within one execution, and no component in the stack holds what an agent learned. There is no named memory layer, no retained preference the agents reapply for a user or requester, no decision trace carried between tickets, no stated scope or lifetime for anything held, and no customer surface to inspect, edit, export or purge agent state apart from tickets, skills and knowledge. Ticket history is a record of work.

Sourcedocs.serval.com self-hosting prerequisites and release notes indexread 2026-09-12

There is no agent memory. One platform holds employee records, assets, permissions and request history so every request carries full context, but that is the customer's application data. It sits under Data Management as people records with employee details, request history and linked assets, an app inventory of SaaS applications with linked people, and an equipment inventory of hardware and its assignees.

These are asset and HR registers, which persist because they are a database, as in a CRM or ERP. Siit says institutional knowledge compounds with each interaction, which describes the corpus filling up as tickets are answered, and it does not say that what compounds can be retrieved, scoped, inspected or tied to a run.

The AI features are exactly three, AI Article Suggestions, Agents and AI Agent Analytics, and no plan item names memory, retained context, agent state or personalization carried between requests.

The agent keeps no requester preference to reapply, no correction that carries to the next ticket and no decision trace across sessions, Siit gives no scope or lifetime for agent state, and there is no way to inspect, edit, export or purge it apart from the people, asset and request records.

Sourcesiit.io/pricing feature matrix, with developer.siit.io resource listread 2026-09-12

Workflows can pause for approval, so execution halts mid flow and waits for a person, and customers author these gates in a redesigned approval flow builder. Around them sit custom approval workflows, natural language workflow approvals, approval delegation, team managers able to override approvals from a ticket, and an approvals API for custom follow up logic. Multi step approvals apply to workflows and access requests alike.

Access requests route through approval before a grant is made, just in time access is time bound and revoked automatically when the window ends, and the access history export records the approver and the justification against every grant. Where approval is not requested, guardrails apply. Serval markets deterministic workflows with security layers, namely RBAC, approval gates, API scope limits and an air gap between help desk and builder.

That air gap means the agent that talks to employees cannot edit workflows or integration scopes, so a compromised conversation cannot widen the agent's own authority. Skills and runbooks hold agent behavior to the team's written procedures, with a diff view for reviewing changes the AI suggests before they take effect.

Sourcedocs.serval.com release notes index and serval.com/insights buyer's guideread 2026-09-12

Approvals are a product object with their own API. On every plan, including the entry tier, a request can trigger validation from a manager or stakeholder, and approvals can also be built into automated workflows. The API can request an approval, list approvals across all requests the user can access, list the approvals for one request and retrieve an approval. The approval sits in front of the consequential action.

Automated app access management lets a customer define who can request each app, routes approvals automatically and provisions access through the identity provider, with an audit trail on every request. The approval stands between the request and the provisioning call, and the decision is recorded against the grant. Time based deprovisioning bounds what an approval can grant, so approved access cannot outlive its window. Siit also describes clean handoffs when the agent escalates, and rapid approvals with fallback rules.

On the Pro tier the agent resolves requests on its own and hands off with context when a request needs judgment, through group assignment to a team inbox, followers who keep stakeholders informed, availability based routing so the handoff reaches someone on shift, and satisfaction ratings on how the request was handled. There is no review step for the AI agent's own replies to employees before they are sent. Approvals cover access provisioning and workflow actions, the operations that cannot be undone.

Sourcesiit.io/pricing feature matrix and developer.siit.io approval endpointsread 2026-09-12

Serval holds SOC 2 Type II, states adherence to HIPAA and GDPR, and documents HIPAA-oriented deployment and data controls. It points customers to its Trust Center for the current Type II attestation report, and a written BAA covers HIPAA workloads and a DPA covers GDPR. Access management is centralized, with SAML and SCIM for SSO and user provisioning, role based access control for least privilege permissions, customizable API scopes for app integrations, and multi step approvals for workflows and access requests.

Further controls include custom roles, scoped API keys, knowledge source permission enforcement, ticket visibility groups for access controlled routing, a read only agent role, default access profiles and app owners, skills gated by access profile, and sharable access reviews. Separation of duties takes the form of an air gap between help desk and builder, so the help desk agent that talks to employees cannot edit workflows or integration scopes or change what agents are allowed to do.

Sourceserval.com/updates/soc2, /insights/hipaa-compliant-itsm and the serval.com platform pageread 2026-09-12

Siit holds SOC 2 Type II and is GDPR compliant, and both link to a Trust Center at trust.siit.io. SOC 2 Type II compliance is listed on every plan, and Siit describes itself as GDPR compliant and SOC 2 Type 2 certified. Siit says it ensures data confidentiality, integrity and availability. Every plan includes role based access control, which assigns roles and permissions, and SSO with Google, Microsoft and Okta using company credentials.

The API enforces the permission model. Request listings are scoped to what the API key owner can see, requests outside that scope return 404 instead of being filtered, and an assignee admin must belong to the assignee inbox. Workspaces, team inboxes and group assignment give the model its shape.

The wider program covers logical tenant separation, encryption in transit at TLS 1.2 and 1.3 and at rest with AES-256, a web application firewall and DDoS protection, regular vulnerability scanning, network and perimeter protection, incident monitoring, vendor risk management, and security education and awareness training. There is a trust center and a public status page. No SCIM provisioning is named, and the SOC 2 report's scope and date are not published.

Sourcesiit.io/security and siit.io/pricing feature matrix and FAQread 2026-09-12

Every workflow run is logged step by step with inputs, outputs, status and duration. Serval offers a sample workflow run export with per-step detail and timestamps and an access history export with approver, justification, start/end and revocation reason, and its governance guidance tells customers to confirm that logs include per-step inputs and outputs, precise timestamps and CSV/JSON export for GRC integration.

An access record carrying the approver and the justification is an audit trail of a consequential agent action with the human decision attached. SIEM integration carries it into the customer's security estate, so evidence is generated when work happens rather than reconstructed before an audit.

Alongside sit agent powered run debugging, version diffs for workflows and skills, a guidance diff view for reviewing AI suggestions, automation suggestion history and workflow journeys viewable as dependency graphs, so a reader can see what ran, what changed and what the agent proposed.

Dashboards on request volume, time to resolution and SLA compliance measure the help desk operation, AI resolution rates show what the agent itself did, and the Insights Agent reports on ticket patterns and automation opportunities. Analytics dashboards are in public beta, with a refreshed home.

Sourceserval.com/insights SOC 2 governance page, with docs.serval.com release notesread 2026-09-12

Two things show what the agent did. AI Agent Analytics, its own line item in the AI group and separate from the request dashboards, gives real time and historical insight into IT agent performance. An audit trail covers the agent's most consequential work, since automated app access management keeps an audit trail on every request. Access grants and revocations are the irreversible things Siit does, and each carries a record.

Through the API the trail can be read and attributed: approvals can be retrieved for each request, messages and private notes are listed, status and assignee changes have their own endpoints, and archive and unarchive are recorded transitions. Data exports download dashboard reports in several formats to share or analyze offline, beside an Analytics Dashboard covering requests, assets and performance. A public status page at status.siit.io covers service availability.

Enterprise grade reporting tracks resolution rates, costs, SLA performance and bottlenecks, and every service request is logged. Those measure the customer's support operation, not the agent. AI Agent Analytics sits on the Pro plan alongside the agents it measures. Siit does not publish the audit trail's fields, retention or detail for individual agent runs.

Sourcesiit.io/pricing feature matrix, with developer.siit.io API referenceread 2026-09-12

Serval comes in four delivery models. Fully managed Serval Cloud has no infrastructure to manage, automatic updates and a 99.9% uptime SLA. In Hybrid Self-Hosted, the customer runs only the Serval worker inside its own infrastructure while Serval operates the backend, control plane, web app and database, so the worker is the only Serval component that ever touches the internal network or handles integration credentials.

Fully self hosted comes in two operating modes, Managed in Your AWS and Self-Managed on Your Kubernetes. It is single tenant in an AWS account the customer owns, the application runtime and all customer data remain in that account, and the customer's AWS credentials never leave its environment. Either the customer's team or Serval operates it, and the deployment is identical either way.

The deployment bundle holds Terraform, Kubernetes and Helm configuration and setup scripts, and the AWS stack is EKS, Aurora PostgreSQL, ElastiCache, OpenSearch, S3 and Secrets Manager. Customers can point at their own PostgreSQL, Redis, Temporal, OpenSearch or S3 compatible storage, with the upgrade cost of doing so spelled out, and DNS requirements go down to the app and API hostnames.

Private network patterns include Site-to-Site VPN, Direct Connect, Transit Gateway and private DNS resolution. The cloud product also offers a region choice, with EU data residency available to organizations, and Microsoft GCC High is supported for US government workloads. Installation and day two operations sit in a password protected area shared with customers and evaluators, while everything needed to evaluate the models is public.

Sourcedocs.serval.com self-hosting section and release notesread 2026-09-12

Delivery is SaaS on AWS, with no region named and no hosting choice. Siit runs on AWS with network and perimeter protection. It names no region, residency choice or customer hosted option on any plan, including the Custom enterprise tier.

There is no self hosting, hybrid worker, single tenant deployment or bring your own cloud, and Siit names logical tenant separation, which describes a multi tenant architecture.

Siit S.A.S is a French company selling GDPR compliance to European customers, including named EU customers, and it publishes no EU hosting commitment, so a customer with a data localization requirement has nothing to work from.

Sourcesiit.io/security and siit.io/pricing feature matrixread 2026-09-12

Three named agents ship, each doing a different job. The Help Desk Agent handles intake and resolution across Slack, Teams, email, phone and portal, the Automation Agent turns natural language descriptions into TypeScript, and the Insights Agent detects patterns and suggests automations. Without the Insights Agent the help desk still resolves tickets, and without the Automation Agent the other two still work.

A browsable catalog of installable workflow bundles sits alongside them, with packs for specific estates such as LDAP/AD, Okta and Google Workspace, plus workflow backed catalog items and service catalog ingestion. An installable Okta workflow does work when selected, and removing it leaves the others whole. Workflows can be organized into folders, shared across teams, searched semantically and tagged. A native knowledge base and Skills add reusable content, though the customer writes those. The catalog comes from Serval, with no third party or community contribution path and no marketplace of externally built agents.

Sourceserval.com/insights buyer's guide, first party about Serval, with docs.serval.com release notesread 2026-09-12

Siit ships reusable structures that the customer fills. A service catalog, a structured list of services for employees with categories that help admins manage requests, backs a self service portal where employees browse the catalog and submit requests. Service Forms gather the right details for each type of request. Three ITIL request types (Incident, Change and Problem) ship as structured models instead of blank tickets, and the API can create and update Services. The catalog holds the customer's own services, forms hold fields the customer defines, and the request types are process shapes, not working units.

Siit publishes no ready to run content of its own, with no library of prebuilt workflows, no starter automations for common IT tasks and no agent presets. Agents are one capability on the Pro plan, not a set to choose from, and there is no catalog of prebuilt agents. The five workflow families (Equipment, Application, Date, People and Request) are listed separately, but Siit does not say whether they are prebuilt bundles a customer installs or trigger categories a customer builds within.

Sourcesiit.io/pricing feature matrixread 2026-09-12

Serval uses two model providers and splits the work between them, and the customer does not choose. In self hosted deployments, the customer provides its own keys for the providers Serval uses at runtime, an Anthropic API key as the primary provider for AI agents and help desk functionality and an OpenAI API key for embeddings, image description and specific model tasks. In Managed in Your AWS, the customer owns the AI provider keys and any customer-owned credentials supplied during onboarding, and an in product surface manages Anthropic usage.

The customer therefore owns the provider relationship, sees the spend and controls the data path to the model, but that is control over credentials, cost and data path, not model selection. Serval decides which provider handles which task, and there is no model picker, per agent model assignment, version pin or way to substitute another provider for either role. The key arrangement applies to the self hosted and hybrid models, and nothing says a Serval Cloud customer supplies keys.

Sourcedocs.serval.com self-hosting overview and prerequisitesread 2026-09-12

The assistant is AI powered, but the underlying model is not disclosed and the customer has no choice of model. Siit names AWS as its cloud provider and names no model provider or AI subprocessor. No plan, including the Custom enterprise tier, offers a model, provider, routing or bring your own key option, and the AI features are three, AI Article Suggestions, Agents and AI Agent Analytics.

Siit connects to MCP compatible AI tools like Claude or Cursor, but that is the customer's own assistant calling Siit, and choosing that assistant does not choose the model that runs Siit's triage, article suggestions or agent resolution. Siit does not say it routes across more than one model internally either.

Sourcesiit.io/pricing feature matrix and siit.io/securityread 2026-09-12

A large API lets outside callers reach Serval. The documentation index at docs.serval.com/llms.txt lists three sections, Documentation at 142 pages, Integrations at 164 pages and an API reference at 252 pages, and two OpenAPI documents, auth-openapi-spec.yaml and openapi-spec-v2.yaml, are served from the API reference section. Scoped API keys and customizable API scopes for app integrations grant access per integration rather than all or nothing.

An approvals API handles custom follow up logic, and access can be revoked via API and workflows, so an outside caller can drive a consequential action inside Serval. Beyond the REST API there is a CLI, which can raise access requests, and user MCP, which puts Serval inside the customer's own assistant. Webhook triggered workflows cover the event side, and workflows, guidance and configuration are managed as code in Git repositories.

Sourcedocs.serval.com llms.txt index, OpenAPI spec listings and release notesread 2026-09-12

Siit offers a public, versioned REST API and an MCP server. The base URL is api.siit.io with a /v1 suffix, resources carry prefixed UIDs, the rate limit is 60 requests per minute with X-SiitApi-Limit, X-SiitApi-Remaining and X-SiitApi-Reset response headers, and Authentication, Pagination and Errors each have their own page. An OpenAPI specification is published at /api-reference/openapi.yml. The API reads and writes, and the writes reach consequential actions.

Full CRUD covers Applications, Articles, Departments, Equipment, Legal Entities, Office Locations, Requests, Services, Teams and Users. Add a User to an Application and Remove a User from an Application provision and deprovision application access, and a user stays listed while their app_access origin carries status waiting_deprovisioning. Alongside them are Update status, Update assignees, Send a message, Send a private note, Request an approval, Archive and Merge User, and Publish, Unpublish and Archive an Article.

An outside caller can grant someone access to a SaaS application through this API. The MCP server has its own documentation page, and MCP connectivity is a plan line item that connects Siit to MCP compatible AI tools like Claude or Cursor, with access scoped to the features and permissions in the customer's plan. The API is available during the trial and on the Standard and Pro plans, Open API Access is a tiered entitlement, and the 14 day free trial reaches it without a card.

Sourcedeveloper.siit.io API reference and llms.txt index, with siit.io/pricingread 2026-09-12

Help Desk Agent evaluations are in beta, a harness aimed at the agent's own behavior. What an evaluation produces is not stated, with no score, pass rate, judge verdict or comparison format. Workflows and skills are managed as code in Git with version diffs, so a change can be reviewed before it lands.

Agent powered run debugging lets a failed run be inspected, and a guidance diff view for reviewing AI suggestions gates the agent's own proposed changes to its instructions. Serval markets deterministic workflows, so behavior repeats between runs.

Its commercial claim is a guaranteed automation rate, citing customer automation percentages, which measures how much of the customer's ticket volume the agent cleared.

Sourcedocs.serval.com release notes indexread 2026-09-12

Agent performance is measured, but nothing tests a change before it goes live. AI Agent Analytics, a Pro plan entitlement, gives real time and historical insight into how the IT agent performs. Satisfaction ratings collect feedback on request handling, and knowledge gaps are found automatically as the AI learns which questions lack good answers, which shows where the agent falls short.

There is no test environment, sandbox, dry run, holdout, A/B comparison, evaluator or scored verdict on agent behavior, and no versioning of agent configuration or workflow definitions. The API returns workflows read only, with no create, update or version endpoint. A change to a workflow or the knowledge base takes effect at once, with no gate before it goes live and no readable comparison of before and after. There is no formal agent testing framework.

Sourcesiit.io/pricing feature matrix, with developer.siit.io workflow endpointsread 2026-09-12

Integrations and APIs are how Serval acts, not driving a browser or a computer interface, and every connector works through an API. For RSA SecurID, Serval stores a client key and attaches it automatically to every call it makes on the customer's behalf, and for Presidio it exchanges credentials for a short lived JWT and sends it as a bearer token on every request. Provisioning runs via SCIM, direct API or custom workflows. Across 164 connectors the pattern is authenticated service to service calls, and the self hosted worker exists to reach internal APIs, LDAP and Active Directory over the network, not to operate a desktop.

Serval markets deterministic workflows compiled to TypeScript. It integrates Jamf and Tanium and runs device actions across managed endpoints, with every grant, device action and workflow run exportable, and those actions go through the MDM and endpoint platforms' own APIs. The CLI, which can raise access requests, is a way for people to reach Serval, not Serval operating anything. There is no hosted or local browser, desktop session, remote control, RPA or recorder.

Sourcedocs.serval.com RSA SecurID and Presidio integration pages, with the release notes indexread 2026-09-12

Siit acts through system integrations and power actions, not by driving a browser or a computer. It always works through an API. Access is provisioned through the identity provider, app inventories fill by connecting Google, Okta, Entra ID or JumpCloud, and Siit's own API can add a user to an application and remove one. Across more than a hundred connectors the pattern is authenticated service to service calls, and outbound events go out as webhooks.

Nothing in the product reads a rendered page, so a connected SaaS app changing its interface does not affect it. There is no hosted or local browser, desktop session, remote or local computer control, RPA or recorder. The equipment inventory tracks devices as records and does not operate them, and app discovery reads identity provider directories. The Slack and Teams bots work inside another vendor's chat client as first party apps on documented platform APIs, invited by the customer and speaking through published endpoints.

Sourcesiit.io/pricing feature matrix, with developer.siit.io API reference and siit.io/securityread 2026-09-12