MCP Assistant Access | Solvren Docs

Overview

Solvren's MCP server lets approved AI assistants answer questions and help teams act on revenue protection data without giving the assistant broad product access.

Use MCP when you want an assistant to answer questions like:

"What revenue is at risk right now?"
"Which decisions need approval?"
"What proof is missing before this can ship?"
"What should I tell the CEO?"
"Can this assistant prepare an approval or request more proof?"
"What has the assistant done recently?"

The core product sentence still applies:

Solvren shows what revenue is at risk, what needs action, and what value was protected.

MCP exposes that same value loop to trusted assistants.

Who MCP is for

MCP is primarily for enterprise teams that want Solvren context available in approved AI workflows.

Common users include:

| User | What they can do with MCP | |------|----------------------------| | CEO or executive assistant | Ask for a quick summary of money at risk, pending decisions, and protected value | | VP Finance | Review revenue exposure, pending proof, and board-ready summaries | | Engineering VP | Understand what is blocking a release or which revenue-impacting decisions need action | | Risk or compliance manager | Review audit history, policy boundaries, and trust reports | | Admin | Configure assistant access profiles and review controlled-action history |

The interface and vocabulary stay consistent. MCP does not create a separate product model; it gives assistants a governed way to use Solvren's existing Home, Decisions, Problems, Proof, and Setup concepts.

What MCP can do

Solvren MCP is organized into four capability groups.

1. Read Solvren context

Read-only tools help assistants summarize:

Home summary
Decisions needing action
Decision details
Open problems
Problem details
Proof and value summaries
Setup health
Data freshness

These tools are useful for executive briefings, daily reviews, Slack summaries, and internal copilots.

2. Explain what should happen next

Guidance tools help answer:

Is this decision ready to approve?
What proof is missing?
Who still needs to decide?
What is the recommended next step?
What caused this problem?
Which decisions or incidents are related?
What setup gap should be fixed next?

These tools are read-only. They help people understand what matters without changing Solvren data.

3. Draft messages and proof packets

Draft tools can produce:

Approval comments
Requests for missing proof
Owner follow-up messages
Executive updates
Board briefs
Decision briefs
Problem briefs
Setup gap summaries

Draft tools do not mutate data. They return suggested text that a person can review before sending or using.

4. Prepare and confirm controlled actions

When explicitly enabled, MCP can use a two-step controlled action flow:

Prepare an action. The assistant creates a short-lived action request that says exactly what would change.
Confirm the action. A user must provide the exact confirmation phrase before Solvren executes the action.

Controlled actions are intentionally narrow:

Approve a decision
Request more proof
Attach a proof link
Assign a problem owner
Add a problem comment

MCP cannot invite users, change billing, alter licenses, edit integrations, change its own policy, bypass permissions, or perform destructive administration.

Safety model

MCP is designed around the principle that assistants should be useful before they are powerful.

Effective access is the intersection of:

server environment guardrails
customer policy profile
organization allowlist
tool allowlist
mutation allowlist
user role and email domain rules
Solvren object-level permissions
explicit confirmation
audit logging

If any layer blocks an action, the action is blocked.

Access modes

Solvren MCP supports two deployment modes.

| Mode | Meaning | Best for | |------|---------|----------| | readonly | Assistants can read, explain, draft, and generate summaries, but cannot mutate Solvren data | Executive summaries, Slack bots, analyst workflows, secure pilots | | controlled_write | Assistants can prepare approved action types and execute them only after explicit confirmation | Mature enterprise deployments with admin approval |

The default mode is readonly.

Policy profiles

Policy profiles let enterprise admins scope assistant access by deployment.

For example, you might create:

an executive assistant profile that can summarize Home, Proof, and urgent Decisions
a finance assistant profile that can inspect revenue exposure and proof packets
an engineering assistant profile that can inspect Decisions, Problems, and setup gaps
a controlled action profile for a restricted operational assistant

Each profile can define:

| Setting | Purpose | |---------|---------| | Profile key | Which deployed assistant uses this profile | | Allowed tools | Which MCP tools can run | | Allowed actions | Which controlled write actions can be prepared | | Allowed roles | Which Solvren roles may use the assistant | | Allowed email domains | Which user email domains are accepted | | Result limit | Maximum list size returned to the assistant | | Confirmation requirement | Whether controlled actions require explicit confirmation | | Action expiration | How long prepared actions remain valid | | Redaction level | How aggressively sensitive fields are hidden | | Audit level | How much MCP usage metadata is retained |

Admins can review profiles in Solvren at:

Setup -> Organization setup -> MCP assistant access

or directly:

/settings/mcp

Deployment overview

MCP is normally deployed by an enterprise admin, platform team, or Solvren account team.

Required environment

The MCP server runs outside the browser, so it uses trusted server credentials.

Minimum environment:

NEXT_PUBLIC_SUPABASE_URL=...
SUPABASE_SERVICE_ROLE_KEY=...
SOLVREN_MCP_MODE=readonly
SOLVREN_MCP_CALLER_LABEL=executive-assistant
SOLVREN_MCP_POLICY_PROFILE_KEY=executive-assistant
SOLVREN_MCP_AUDIT_ENABLED=true

Common guardrails:

SOLVREN_MCP_ALLOWED_ORG_IDS=org-id-1,org-id-2
SOLVREN_MCP_ALLOWED_TOOLS=get_home_summary,list_decisions_needing_action,get_decision,get_proof_summary
SOLVREN_MCP_MAX_LIMIT=25

Controlled write guardrails:

SOLVREN_MCP_MODE=controlled_write
SOLVREN_MCP_ALLOWED_MUTATIONS=approve_decision,request_more_proof,attach_proof_link,assign_problem_owner,add_problem_comment,confirm_action
SOLVREN_MCP_REQUIRE_CONFIRMATION=true
SOLVREN_MCP_WRITE_AUDIT_REQUIRED=true
SOLVREN_MCP_ACTION_TTL_MINUTES=10

Use the smallest practical allowlist for each assistant deployment.

Example MCP client configuration

{
  "mcpServers": {
    "solvren": {
      "command": "npm",
      "args": ["run", "mcp:server"],
      "cwd": "C:\\Users\\your-user\\revenueguard",
      "env": {
        "NEXT_PUBLIC_SUPABASE_URL": "https://your-project.supabase.co",
        "SUPABASE_SERVICE_ROLE_KEY": "your-service-role-key",
        "SOLVREN_MCP_MODE": "readonly",
        "SOLVREN_MCP_ALLOWED_ORG_IDS": "org-id-1",
        "SOLVREN_MCP_ALLOWED_TOOLS": "get_home_summary,list_decisions_needing_action,get_decision,get_proof_summary,get_setup_health,generate_board_brief,draft_exec_update",
        "SOLVREN_MCP_MAX_LIMIT": "25",
        "SOLVREN_MCP_CALLER_LABEL": "executive-assistant",
        "SOLVREN_MCP_POLICY_PROFILE_KEY": "executive-assistant",
        "SOLVREN_MCP_AUDIT_ENABLED": "true"
      }
    }
  }
}

Do not expose the stdio MCP server over HTTP unless your organization adds authentication, tenant authorization, rate limits, network controls, and audit logging.

Tool reference

Core read tools

| Tool | Use it for | |------|------------| | get_home_summary | Executive summary of money at risk, urgent work, problems, protected value, and setup health | | list_decisions_needing_action | Decisions that need review or completion | | get_decision | One decision with impact, proof, approvals, and next action | | list_open_problems | Open revenue problems ordered by priority | | get_problem | One problem with impact, owner, actions, and sources | | get_proof_summary | Protected value, prevented incidents, saved time, and proof packets | | generate_executive_brief | Short executive brief from current risk, action, setup, and proof | | get_setup_health | Connected systems, setup gaps, and latest scan | | get_mcp_server_health | MCP server mode, audit status, and configured guardrails | | get_mcp_data_freshness | Latest timestamps for Solvren data |

Decision readiness tools

| Tool | Use it for | |------|------------| | explain_decision_readiness | Whether a decision can be approved | | list_missing_proof | Required proof still missing | | list_pending_approvers | Who still needs to decide | | recommend_decision_next_step | The next best action for a decision |

Problem resolution tools

| Tool | Use it for | |------|------------| | recommend_problem_actions | Practical next actions for a problem | | summarize_problem_root_cause | Likely root cause and source context | | list_related_decisions | Decisions linked to a problem | | list_related_incidents | Incidents related to a problem |

Proof and setup tools

| Tool | Use it for | |------|------------| | generate_board_brief | Board-ready Markdown summary | | generate_decision_brief | Decision-specific brief | | generate_problem_brief | Problem-specific brief | | generate_setup_gap_summary | Setup gap summary | | recommend_next_setup_step | The next setup action | | explain_coverage_gap | Why a setup gap matters | | list_unprotected_revenue_surfaces | Revenue areas that appear under-protected |

Draft-only tools

| Tool | Use it for | |------|------------| | draft_approval_comment | Suggested approval or not-ready comment | | draft_request_for_proof | Suggested proof request | | draft_owner_message | Suggested owner follow-up | | draft_exec_update | Suggested executive update |

Controlled action tools

| Tool | Use it for | |------|------------| | prepare_approve_decision | Prepare an approval | | prepare_request_more_proof | Prepare a visible proof request | | prepare_attach_proof_link | Prepare attaching a proof link | | prepare_assign_problem_owner | Prepare assigning a problem owner | | prepare_add_problem_comment | Prepare adding a problem comment | | confirm_action | Execute a prepared action after exact confirmation |

Policy and trust tools

| Tool | Use it for | |------|------------| | get_mcp_policy | Show the effective policy profile | | explain_mcp_policy | Explain the policy in plain English | | list_mcp_policy_profiles | List assistant profiles for an organization | | get_effective_mcp_permissions | Show environment, policy, and actor boundaries | | simulate_mcp_tool_call | Test whether a tool call would be allowed | | simulate_mcp_action | Test whether a controlled action would be allowed | | explain_why_mcp_action_blocked | Explain a blocked controlled action | | get_mcp_audit_summary | Summarize recent MCP action audit | | get_mcp_action_history | List recent MCP action events | | generate_mcp_trust_report | Generate a security-reviewable trust report |

Response format

Every tool returns a stable JSON envelope:

{
  "schemaVersion": "solvren.mcp.v1.2",
  "toolName": "get_home_summary",
  "orgId": "organization-id-or-null",
  "generatedAt": "2026-06-25T00:00:00.000Z",
  "whatThisMeans": "Plain-English summary.",
  "whyItMatters": "Business reason this matters.",
  "recommendedNextAction": "The next sensible action.",
  "sourceRecords": [
    { "table": "change_events", "id": "record-id", "label": "Record label" }
  ],
  "links": [
    { "label": "Open Solvren", "url": "/" }
  ],
  "data": {}
}

The assistant should prefer whatThisMeans, whyItMatters, and recommendedNextAction for executive-facing answers, then use data and sourceRecords for detail and traceability.

Example prompts

Use prompts like these with an approved MCP client:

Give me the Solvren home summary for Bluepeak. What needs attention today?

Which decisions need approval, and what proof is missing?

Generate a board brief for the last 30 days of protected revenue.

Can this assistant prepare an approval action for this decision? Simulate it first.

Generate an MCP trust report for security review.

Admin review checklist

Before enabling MCP for production:

Confirm the MCP server runs only in a trusted environment.
Use readonly mode first.
Restrict allowed organizations.
Restrict allowed tools.
Set a conservative result limit.
Create a customer policy profile.
Require confirmation for controlled actions.
Require MCP audit logging.
Review /settings/mcp.
Generate an MCP trust report.
Test blocked actions before enabling controlled writes.

Troubleshooting

| Symptom | Likely cause | What to do | |---------|--------------|------------| | "Organization not found" | The org ID is wrong or not available to the server | Confirm the org ID and environment | | "This server is not authorized for that organization" | The org is outside the allowlist | Update SOLVREN_MCP_ALLOWED_ORG_IDS or use the correct deployment | | "Controlled writes are disabled" | Server is in readonly mode | Use read/draft tools or intentionally enable controlled_write | | "Tool is not allowed" | Environment or customer policy blocks the tool | Review allowed tools and policy profile | | "Mutation is not allowed" | The controlled action is not allowlisted | Review allowed mutations and customer policy | | "Confirmation phrase does not match" | The user did not provide the exact phrase | Use the phrase from the prepare response | | No policy profiles appear in settings | No database profile exists for this assistant key | Create a profile or rely on environment guardrails |