Revenue Contract Agent is in early access.
Meridian

Platform

The layer under the agents.

One governed layer under every agent you run, whether Meridian built it, a partner built it, or your own team did.

Assist
The conversational front door across Meridian and the systems you already run.
Registry
The system of record for every agent: owner, scope, permissions, and compliance.
Studio
A low-code builder for your own agents, with approvals and evaluations built in.
Gateway
Third-party and custom agents connected through MCP, A2A, and OpenTelemetry.
Data Fabric
Zero-copy reads from your warehouse, plus a lakehouse and 3,000 connectors.
Trust
Human approval on consequential actions, an immutable audit trail, your model.
Meridian platform architectureAssist is the front door for employees, managers, and finance teams. It runs agents that are recorded in the Registry, the system of record. The Gateway sits at the edge of the trust boundary and connects third-party agents over Model Context Protocol and agent-to-agent protocols, identity providers over OpenID Connect, SAML, and SCIM, and observability tools over OpenTelemetry. Studio publishes custom agents into the Registry. The Data Fabric at the base reads Snowflake, Databricks, BigQuery, an Apache Iceberg lakehouse, and 3,000+ connectors in place. Trust bounds all of it with approval holds, an immutable audit trail, role-based access, and data residency.Trust boundaryEmployees · managers · finance teamsAssistFront door · browser, Slack, TeamsRegistrySystem of record for every agentOwner · role · permissions · data · statusStudioLow-code builderApprovals built inEvaluation setsGatewayPolicy on every callMCP · A2A · OTLPOIDC · SAML · SCIMData FabricZero-copy · security inheritedSnowflakeDatabricksBigQueryApache Iceberg3,000+ connectorsApproval holds · audit trail · RBACSOC 2 Type II · ISO 27001 · EU/US residency
Six parts, one boundary. Assist is the front door, Registry the system of record, Studio the builder, Gateway the policy point at the edge, Data Fabric the base, and Trust the constraints every agent runs inside.

RegistryThe system of record for agents

Every agent in your enterprise. One record each.

One registry of every agent, whether built by Meridian, a partner, or your own team, with owner, role, permissions, data touched, and compliance status. Workforce analytics that count agents and people together.

  • One record per agent

    Owner, role, purpose, permissions, model, data scopes, and status for every agent, including partner and customer-built ones registered through the Gateway.

  • Permissions and data lineage

    Agents hold no standing credentials. Each inherits scopes from your identity provider and the source system's security model. The Registry shows exactly which tables and documents each agent can read.

  • Compliance status

    Review dates, risk tier, approved use cases, and evaluation results per agent. Agents that miss a review are flagged; agents that fail an evaluation can be paused automatically.

Explore Registry
app.meridian.example/registry
24 agents
AgentsPeopleFilter
AgentOwnerCompliance
Payroll AgentM. ReyesCompliant
Help Desk AgentD. OkaforCompliant
Audit AgentK. LindqvistCompliant
Controls AgentS. BhattReview due
Close AgentT. NakamuraCompliant
Vendor Risk CheckP. AdeyemiPaused
Expense AuditorL. MarshCompliant
Scheduling AgentR. CastilloCompliant

GatewayOpen standards, governed

Bring any agent. Govern it here.

The Gateway connects third-party and custom agents to the Registry through open standards: Model Context Protocol for tools, agent-to-agent protocols for handoffs, OpenTelemetry for observability, and your identity provider for access.

  • MCP tool servers

    Expose your systems as Model Context Protocol tools with per-agent scopes, schema validation, and rate limits. Meridian ships servers for common HR and finance systems; add your own.

  • Agent-to-agent handoffs

    Agents delegate sub-tasks to other agents over A2A-style protocols. The originating user's identity and approval context travel with the request.

  • Policy enforcement point

    Every tool call is checked against policy: allowed tools, data scopes, cost ceilings, and approval requirements. Denials and holds are logged with the rule that fired.

Explore Gateway
app.meridian.example/gateway
Policy v41
Connected agents
17
Calls today
48,211
AgentProtocol
Planning AgentMeridianMCP
Contract Review AgentMeridianMCP
Expense AuditorPartnerA2A
Vendor Risk CheckStudioMCP
Treasury ForecastCustom · PythonA2A
Ticket TriagePartnerMCP
Held for approvalControls Agent → ap.release_payment$48,211.90 · INV-2291awaiting S. Bhatt · rule AP-14 · policy v41

Data FabricZero-copy access

Agents read your data where it lives.

Zero-copy access to Snowflake, Databricks, and BigQuery, governed SQL, an Apache Iceberg lakehouse, and more than 3,000 prebuilt connectors. Row and column security is inherited, not reimplemented.

  • Zero-copy warehouse access

    Agents query Snowflake, Databricks, and BigQuery in place through native sharing. No extract, no second copy, no drift.

  • Inherited security

    Row-level and column-level policies defined in the warehouse or source system apply to every agent query. Agents cannot see more than the user they act for.

  • Apache Iceberg lakehouse

    An open-format lakehouse you own, for sources without a warehouse footprint. Query it with the engine of your choice.

Explore Data Fabric
app.meridian.example/data
3,012 connectors

Sources

8 connected

  • Snowflake

    Zero-copy

    4 min ago

  • Databricks

    Zero-copy

    9 min ago

  • BigQuery

    Zero-copy

    2 min ago

  • HRIS

    Connector

    12 min ago

Governed SQL

as d.okafor · Snowflake

select cost_center,       sum(actual) - sum(plan) as variancefrom finance.gl_actuals ajoin finance.plan p using (cost_center, period)where period = '2026-Q3'group by 1order by 2 desclimit 5;
Cost centerVariance
Field operations+268,412.55
Cloud infrastructure+144,910.12
Contract labor+98,220.00
Travel+46,114.40
Marketing programs−18,119.25

5 rows · 412 msfreshness: 4 min

StudioLow-code agent builder

Build your own agents on the same guardrails.

Studio is a low-code builder for agents on top of your data and Meridian's controls. Approval steps, evaluation sets, and Registry registration are built in, not bolted on.

  • Visual composition

    Triggers, data reads, reasoning steps, tool calls, and approvals as blocks. Code where you want it, none where you do not.

  • Approval steps as blocks

    Route consequential actions to named approvers or roles. Approvals are enforced at the Gateway, so a builder cannot skip them.

  • Evaluation sets

    Define expected behavior on real examples. See pass rates and regressions before publishing and on every model change.

Explore Studio
app.meridian.example/studio/ap-duplicate-check
Draft v7

AP duplicate check

5 blocks · autosaved 14:02

  1. Trigger01

    Invoice posted

    ERP event · AP subledger

  2. Read02

    Vendor, PO, history

    Data Fabric · ap:read

  3. Reason03

    Duplicate and policy check

    Workspace model

  4. Approve04

    Hold if over $10,000.00

    AP manager role

  5. Act05

    Release or flag

    ERP via Gateway

Run evaluationPublish

AssistThe conversational front door

Ask once. Search, generate, or run an agent.

Assist is the conversational interface across Meridian and your connected systems. Search with citations, generate documents from governed data, and start agents from the same place, with your permissions applied.

  • Search with citations

    One query across your document stores, HRIS, ERP, and collaboration tools. Every result carries its source and the user's existing permissions apply.

  • Generate from governed data

    Job descriptions, policy summaries, variance narratives, and board-ready tables generated from data the user can already see.

  • Run agents from chat

    Start the Audit Agent on a PBC list or the Scheduling Agent on an open shift from the same conversation. Approvals still apply.

Explore Assist
app.meridian.example/assist
Planning Agent

What drove the Q3 opex variance for Field operations, and which lines are over plan?

Planning Agent · read finance.gl_actuals and finance.plan as d.okafor · Snowflake, 4 min old · 1.8 s

Field operations is $268,412.55 over plan for Q3 2026 (+6.5%). Three lines account for 91% of the variance.

LineVariance
Contract labor+144,300.00
Overtime+66,918.20
Fleet maintenance+33,206.35

gl_actuals · 2026-09-08plan_v3 · FY26

Run: draft commentaryOpen in Planning Agent
Ask about your HR and finance data

Acting as D. Okafor · Northwind Logistics · your permissions apply

Trust

Humans stay accountable. The log proves it.

The same constraints apply to every agent on the platform, whether Meridian built it, a partner did, or you did in Studio.

Security and trust

  • Human-in-the-loop approvals

    Consequential actions are held at the Gateway until a named approver acts. Approval policies are versioned and cannot be bypassed by an agent or a builder.

  • Immutable audit trail

    Every read, reasoning step, tool call, approval, and outcome is logged with the acting identity, hashed, and exportable to your SIEM or archive.

  • Role-based access

    Identity from your IdP over OpenID Connect or SAML, provisioning over SCIM, and source-system permissions inherited on every query.

  • Certifications and regulation

    SOC 2 Type II report and ISO 27001 certification. GDPR compliance with a standard DPA. HIPAA-ready deployment with a BAA on Enterprise.

  • Data residency

    Workspaces in the EU or the US. Data storage, processing, and model inference stay in the region you select.

  • No training on customer data. Your model.

    Customer data is never used to train Meridian's or any third party's models, by contract. Choose frontier models from leading providers; change them per workspace.

  • SOC 2 Type II
  • ISO 27001
  • GDPR
  • HIPAA (BAA on Enterprise)
  • OpenID Connect and SAML
  • SCIM
  • OpenTelemetry (OTLP)

Open standards

Which open standards does the platform use?

Six. Agents connect over Model Context Protocol and agent-to-agent protocols, export traces as OpenTelemetry, take identity from OpenID Connect, SAML, and SCIM, and store connector data in Apache Iceberg. No proprietary protocol is required to join.

StandardWhat it isWhat it means for you
Model Context ProtocolMCPAn open protocol that exposes tools and data to AI agents through typed servers and clients.Any MCP-capable agent can use your systems through the Gateway, with per-agent scopes, schema validation, and rate limits added by Meridian.
Agent-to-agent protocolsA2AOpen protocols for one agent to delegate a task to another agent and receive the result.Meridian, partner, and custom agents hand off work to each other with the originating user's identity and approval context attached.
OpenTelemetryOTLPThe CNCF standard for emitting traces, metrics, and logs in one format.Every agent step exports to the observability platform you already run, so agent traces sit next to application traces.
OpenID Connect and SAMLOIDC / SAMLFederated authentication standards supported by enterprise identity providers.People sign in with your IdP. Agents act as the user and hold no credentials of their own.
SCIMSCIM 2.0The standard for automated user provisioning and deprovisioning between systems.Offboarding a person in your IdP suspends the agents that person owns within minutes.
Apache IcebergIcebergAn open table format for large analytic datasets in a lakehouse.Connector-fed data lands in tables you own and can query with Spark, Trino, DuckDB, or your own warehouse.

Questions

What buyers ask about the platform

Scope, plans, third-party agents, and where data lives.

The Meridian platform is the governed layer under every agent you run. It has six parts: Registry, the system of record for agents; Gateway, which connects agents through open standards and enforces policy on every call; Data Fabric, zero-copy access to your data; Studio, a low-code agent builder; Assist, the conversational front door; and Trust, the approvals, audit trail, access control, certifications, and residency that apply to all of them. The same six apply to agents built by Meridian, by partners, and by your own teams.

No. Meridian's twelve agents use Registry, Gateway, Data Fabric, and Trust by default, so most design partners started there. Studio is for teams building their own agents and Assist is the interface for people; both are optional. Every component shares one identity model, one policy layer, and one log.

Starter includes Assist, approvals, the audit log, and prebuilt connectors for one workspace with up to 3 GA agents. Growth adds the Registry with blended workforce analytics, the Gateway for partner and custom agents, SSO over OpenID Connect or SAML with SCIM provisioning, and EU or US residency. Enterprise adds the Data Fabric with zero-copy warehouse access and an Iceberg lakehouse, Studio, a dedicated environment, and a 99.95% uptime SLA. See pricing for credits and limits.

Yes. Any agent that speaks Model Context Protocol for tools and authenticates with your identity provider registers through the Gateway. Once registered it has a Registry record, the same approval holds, and the same immutable log as Meridian's own agents.

In your warehouse for zero-copy sources (Snowflake, Databricks, BigQuery) and in an Apache Iceberg lakehouse you own for connector-fed sources, in the EU or US region you select. Meridian does not keep a separate copy of your data for its own purposes and does not train on customer data.

Get started

See the platform run on your systems.

A 45-minute working session with your warehouse and identity provider in scope. Leave with an architecture for the first agent and a credit estimate.

Book a demoView pricing