Moterra

Every connector reads and writes, according to Moterra, with agents pulling data and pushing updates across CRM, comms, support, docs, code and data while inheriting the permissions already in place. Its most popular connectors are Salesforce, Slack, Microsoft Teams, SharePoint, Jira and Google Workspace, beside a 100+ figure. The Bridge page adds Text-to-SQL against customer databases and tools built on Dify that reach Salesforce, Jira, Zendesk, Confluence, SAP and custom internal systems through MCP and standard APIs. The full connector list is not published.

Sourcemoterra.ai home and moterra-ai-bridge pagesread 2026-09-29

The vendor's Bridge layer ships Dify Workflows and Agents as a standard plugin hosted inside the customer's VPC, and states that after go-live the customer builds additional agents through Dify's visual workflow builder, with complex workflows using APIs and conditional logic.

Agents can combine knowledge bases, MCP tools, Dify workflows and built-in capabilities, and Claude runs multi-step automations built on those workflows. The customer gets a visual flow surface with conditional logic. The engine is the open-source Dify platform, deployed and operated by Moterra as part of Bridge rather than wired in by the customer.

Sourcemoterra.ai moterra-ai-bridge pageread 2026-09-29

The vendor's permission-aware SharePoint RAG supports multiple knowledge bases, each scoped to specific SharePoint sites, folders or files, with hybrid retrieval using vector-based chunks for large document sets and whole files where they fit the context window, and states the vector stores stay in the customer's AWS environment. Queries impersonate the user's Entra ID identity, so answers are grounded only in documents that user can access. Text-to-SQL adds an automatically built semantic layer over connected databases with schema modeling, view management and query templates.

Sourcemoterra.ai moterra-ai-bridge pageread 2026-09-29

Admins control which models each user can access, set token quotas per person and enforce device policies through MDM with Intune or Jamf, all of which constrain what users and their agents can do. No approval step, checkpoint, review queue or escalation in the run path is described; these are constraints rather than a gate.

Sourcemoterra.ai moterra-ai-bridge and private-claude-cowork pagesread 2026-09-29

Access controls are strong and sit in the customer's own identity system; no attestation of Moterra's own is published.

Sign in runs through the customer's own identity provider with Microsoft Entra ID or Google Workspace SSO, provisioning through MDM with IAM or Entra ID, ACL groups and API tokens managed in the admin console, per database access control, Entra ID impersonation so retrieval honors each user's SharePoint permissions, and customer managed KMS keys with private endpoints.

The product pages badge ISO 27001 and ISO 42001 as aligned, and the home page states that the ISO certificates and SOC 2 Type II reports are held by AWS and Anthropic, with Moterra's own setup independently validated by cyber security specialists and no report named.

Sourcemoterra.ai home, private-claude-cowork and moterra-ai-bridge pagesread 2026-09-29

By default every user prompt, model response, plugin call and administrative action is captured in AWS CloudTrail plus a dedicated log bucket in the customer's AWS account, with retention and access controlled by the customer, and enterprise deployments can forward logs to a SIEM such as Splunk, Sentinel or Datadog or to a custom destination.

That record covers the agent's own execution call by call, rather than activity in the customer's other systems. The admin console adds real time cost tracking per user and per model. The pages do not say how much of the per call record Bridge writes itself rather than inheriting from AWS logging.

Sourcemoterra.ai moterra-ai-bridge and home pagesread 2026-09-29

No agent written memory, conversation persistence or learned context is described on the product pages. Skills that package prompts, tools, knowledge bases and workflows are configuration shipped to users rather than memory, and the vector stores Bridge keeps in the customer's account are retrieval indexes over customer documents. Whether conversations persist between sessions is not published.

Sourcemoterra.ai moterra-ai-bridge and private-claude-cowork pagesread 2026-09-29

The private deployment runs entirely inside the customer's own AWS account, with the customer owning the account and network, the customer managed KMS keys and the choice of region. Any AWS region where Claude on Bedrock is available can be selected, and London eu-west-2, Frankfurt eu-central-1 or Ireland eu-west-1 are recommended for UK and EU customers. The home page adds a Microsoft Azure AI Foundry tenant as the second customer environment, with Google Cloud and on premises deployment available on request. The Bridge layer deploys single tenant into the same AWS account and VPC.

Sourcemoterra.ai private-claude-cowork, home and moterra-ai-bridge pagesread 2026-09-29

Agents are built once by the customer's power users and distributed across teams. During onboarding Moterra builds an initial set of agents for standard use cases such as document analysis, meeting summaries and structured data extraction, and agents can draw on 12 built in capabilities including charts, artifacts, translation, deep research and image generation. The onboarding agents are built per customer as a service rather than drawn from a published catalog, and the built in capability set is only partly named. The named, priced MCP plugins are integration components rather than agents.

Sourcemoterra.ai moterra-ai-bridge and home pagesread 2026-09-29

Users get Claude Cowork's multi step task automation, Claude can run multi step automations built on Dify workflows hosted in the customer's VPC, and existing workflows or automations come up when Moterra describes model deprecation. No scheduled, cron, webhook or external event trigger is named, and no channel beyond the Cowork and Code applications is described, so nothing documented wakes an agent without a user.

Sourcemoterra.ai moterra-ai-bridge, private-claude-cowork and home pagesread 2026-09-29

All models come from one maker. The private deployment offers every Claude model, naming Haiku, Sonnet, Opus and Fable, delivered through AWS Bedrock or Azure AI Foundry, and the Bridge control table lets the customer set LLM and Claude model access per user. Every model named comes from Anthropic, and no other maker's model is named. Bedrock and Azure are two ways to serve those same models, and per user access to one maker's models is not a choice across makers.

Sourcemoterra.ai private-claude-cowork, home and moterra-ai-bridge pagesread 2026-09-29

API tokens are managed in the admin console, which points to an API for the Bridge platform. No API reference, endpoint or authentication documentation is published on the product pages, though a help center is linked. Bridge's four standard MCP plugins, SharePoint RAG, Text-to-SQL, Web Search and Scrape, and Dify Workflows and Agents, expose the customer's own systems and third party tools to Claude.

Sourcemoterra.ai moterra-ai-bridge and home pagesread 2026-09-29

No evaluation, scoring, benchmark, comparison or regression capability is described on the product pages. The vendor states that when Anthropic releases a model, Moterra tests it against the customer's setup and MCP plugins before making it available, which is a service the vendor performs rather than a customer-operated surface. Automated use-case discovery categorizes anonymized conversations into business use cases and reports time saved per case, which measures adoption rather than scoring or comparing agent output.

Sourcemoterra.ai moterra-ai-bridge and private-claude-cowork pagesread 2026-09-29

The vendor's Web Search and Scrape plugin searches, retrieves and scrapes web content for research tasks, which is retrieval rather than operation of software. The Claude Cowork capabilities the vendor lists for its deployment are files, artifacts, code execution, web search, skills and multi-step task automation, none of which is browser control, navigation, form filling or screen interaction, and no such capability is named on the product pages.

Sourcemoterra.ai moterra-ai-bridge and private-claude-cowork pagesread 2026-09-29