Product
Solutions
Resources
Pricing About Security Contact
FRENESDEPTIT
Book a demoLog in
MCP server and API

Your AI agents on your contracts.Under your control.

An MCP server, a REST API, webhooks.Four autonomy modes.Depending on the mode chosen, your agents' actions wait for approval from someone on your team.

app.pactolane.com · MCP · Pending actionsScale plan
Actions pending approvalRequested by your agents, to approve or reject within 24 hours
3 pending
Requested actionRisk level
Create a contract from the “Standard NDA” templateLow

“Sales agent” key · assisted mode

Create a contract from the “Services” templateMedium

“Sales agent” key · assisted mode

Create a contract from the “Subcontracting” templateHigh

“Internal assistant” key · supervised mode

The problem

AI agents are coming.
Your contracts must stay governed.

01

Contracts cut off from your tools

Your teams work in AI assistants and internal applications, but contracts stay out of reach, consulted by hand.

02

All-or-nothing access

Opening contracts to an agent without setting what it can read, write or delete means accepting that it acts without control.

03

Actions with no trail

When a tool modifies a contract, you need to know who requested it, with which key, and who authorized it.

The solution

Programmable access,
governed key by key.

Pactolane's Scale plan opens your contracts to your tools and AI agents through an MCP (Model Context Protocol) server, a REST API and webhooks.Each access key carries its permissions, an expiration date if you wish and an autonomy mode; in supervised mode, any action requested by an agent waits for someone on your team to approve it, and in assisted mode, only write actions wait.

Keys with precise permissions

Each key has a name, permissions chosen from more than twenty (read contracts, create contracts, read templates…) and an optional expiration date.

New key · Sales agent
Permissionscontracts:read, contracts:create
Autonomy modeAssisted
ExpirationMarch 31, 2027
Key displayOnce only

Four autonomy modes

Read-only, supervised, assisted or autonomous: you decide, key by key, how far an agent can act without human approval.

Autonomy modes
Read-only · no changesRead
Supervised · every action approvedApprove all
Assisted · writes approvedApprove writes
Autonomous · direct actionsNo approval

Pending actions

Each action to approve shows the key, the action type, its parameters and a risk level. Without a decision within 24 hours, it expires.

Actions · history
Create a contract · Standard NDAApproved · 10 min ago
Create a contract · SubcontractingRejected · yesterday
Create a contract · ServicesExpired · no decision within 24 h
How it works

Connect an agent
in three steps.

1

You create a key

Name, permissions, autonomy mode and expiration date. The full key is displayed only once: you copy it into your agent or tool.

Key createdActive
Prefix visible in the list
Copy it now
2

Your agent gets to work

Through the MCP server, the agent reads and proposes actions within the limits of its permissions. Your applications use the REST API and receive webhooks.

MCP serverBuilt-in documentation
Contract reading allowed
Writing subject to approval
3

You approve what matters

Actions that require approval arrive on the pending actions page, with their risk level. You approve or reject.

Action · Create a contractMedium risk
ApproveReject
Before & After Pactolane

From manual exports
to governed access.

SituationWithout PactolaneWith Pactolane
Giving contracts to an AI agentCopying documents into a conversationAn MCP server with per-key permissions
Limiting what it can doTrust it or block everythingFour autonomy modes, from read-only to autonomous
Authorizing a sensitive actionAn informal exchange, with no trailA pending action, approved or rejected, with its risk level
Connecting your applicationsPeriodic exports and importsREST API and webhooks, with an integration log
Cutting off accessChanging a shared passwordRevoking the key in question, without touching the others
Features

Open up your contracts,
without losing control.

MCP server

The MCP protocol lets external tools and AI agents access Pactolane data in a secure, controlled way.

REST API

API keys managed by the administrator to connect Pactolane to your applications, with dedicated documentation.

Webhooks

Webhook creation, test sends and an integration log to track every exchange.

Per-key permissions

More than twenty permissions to select, an expiration date, and revocation or deletion at any time.

Human approval

Pending actions filterable by status, with a low, medium, high or critical risk level, and expiration after 24 hours.

Built-in documentation

Overview, architecture, key management, usage guide and security, in the MCP tab of the Integrations page.

In practice

Clear, verifiable
safeguards.

4
Autonomy modes: read-only, supervised, assisted, autonomous.
4
Action risk levels: low, medium, high, critical.
24h
Time after which an unhandled action expires.
Security

Open up access
without lowering security.

Access security, key by key.

  • Permissions chosen one by one for each key, with an optional expiration date
  • MCP keys displayed only once, revocable or deletable at any time
  • One autonomy mode per key, and human approval of actions according to that mode
  • A risk level on every pending action, which expires after 24 hours without a decision
  • SSO and advanced audit trail to trace every access and every decision
  • Data hosted in France and Belgium, on Google Cloud, encrypted at rest with AES-256-GCM
Per-key permissionsAutonomy modesSSOAdvanced audit trailFrance and BelgiumGDPR by default
FAQ

Common questions.

Q.01What is an MCP server for contract management?

An MCP (Model Context Protocol) server is a standard gateway that lets an AI agent or an external tool view data and request actions in an application. In Pactolane, it opens your contracts and templates to your agents, within the limits of each key's permissions and, depending on the autonomy mode chosen, with human approval of all actions or of write actions only.

Q.02How does Pactolane control what an AI agent does?

Pactolane controls each agent through its access key: permissions chosen one by one, an optional expiration date and an autonomy mode. In read-only mode, the agent changes nothing; in supervised mode, every action waits for human approval; in assisted mode, reads are direct and writes wait for approval; in autonomous mode, actions run directly. Pending actions show their risk level and expire after 24 hours without a decision.

Q.03What's the difference between the MCP server and the REST API?

In Pactolane, the MCP server is designed for AI agents, with its keys, its autonomy modes and its queue of actions to approve. The REST API and webhooks are used to connect Pactolane to your applications: your administrators generate API keys, create webhooks, test them and track exchanges in the integration log. Both are included in the Scale plan.

Q.04How do I secure an MCP access key?

In Pactolane, the full MCP key is displayed only once, when it is created: you immediately copy it into a vault or into your agent's configuration. The key list then shows only its prefix, its status and its last use. You can revoke or delete it at any time, and give it an expiration date as soon as it is created.

Q.05Which plan includes the MCP server and the API?

Pactolane's MCP server, REST API and webhooks are reserved for the Scale plan (from €2,500 per month excl. VAT, custom quote), which also includes SSO, the advanced audit trail, dedicated support and legal onboarding. Full details are on our pricing page, and the other connectors on our integrations page.

Connect your agents while staying in control.

The Scale plan is built with you: let's talk about your use cases.

Contract risk gives no warning. Your watch does.

Every week, field insights on contracts, risks and best practices.
For legal, procurement and IT leaders.

FreeOne email per weekUnsubscribe in one click

By subscribing, you agree to our privacy policy.

Cookies & privacy

Pactolane uses analytics cookies to understand how you use this site and improve its content. No personal data is ever sold or used for advertising. Learn more about our cookie policy