Continua AI

Continua MCP Server gives any MCP-compatible client the full investigation engine: cited analysis across your documents, Office and data deliverables, and projects that accumulate reusable knowledge, decisions, and rules.

Hosted MCP Server

npx add-mcp 'https://app.askcontinua.com/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

Continua MCP for Claude

Continua AI

Continua AI brings cited investigations, reusable project knowledge, and working deliverables into Claude. Use it for private credit, factoring, supply chain finance, and enterprise knowledge management: start with financials, agreements, invoices, correspondence, and other source materials; return with a credit memo, exposure table, or review pack your team can act on.

This Claude Code plugin includes the /continua:analyze skill, connects to the hosted Continua MCP server, and supports Investigation, Project Room, and Automation workflows.

Follows the Claude Code plugins guide and plugin publishing guide.

Connect Continua after installation

In Claude, open Settings → Connectors, find Continua, enable it, and choose Connect to sign in. In Claude Code, run /mcp, select the Continua server, and complete authentication. If authorization expires, reconnect there and retry the request.

The plugin uses Streamable HTTP with OAuth:

https://app.askcontinua.com/mcp

Before the first task, call learn_how_to_use with {}. It returns the current usage guide without creating an analysis or consuming analysis credits. Investigations and project searches may consume credits; check the account's billing information when needed.

For clients using API keys, follow the MCP connection guide. Keep credentials in your client's secure configuration.

Plugin layout

continua-mcp/                  # Plugin root
├── .claude-plugin/
│   └── plugin.json            # Manifest only
├── .mcp.json                  # MCP: Streamable HTTP + OAuth
├── logo.png                   # Plugin icon
├── LICENSE                    # MIT license
├── skills/
│   └── analyze/
│       └── SKILL.md           # /continua:analyze
└── README.md
FileRole
.claude-plugin/plugin.jsonPlugin identity (name: continua), version, and MCP configuration
.mcp.jsonConnection to https://app.askcontinua.com/mcp
logo.pngPlugin icon
LICENSEMIT license
skills/analyze/SKILL.mdTool picker and Investigation, Project Room, and Automation workflows

Only plugin.json belongs in .claude-plugin/; keep skills/, .mcp.json, and the logo at the repository root.

Plugin icon: Continua logo. Privacy policy: Continua privacy policy.


Local development

From this repository's root:

claude --plugin-dir .

In Claude Code:

  1. Enable the plugin if prompted.
  2. Run /mcp, connect continua, and complete sign-in.
  3. Run /reload-plugins after editing SKILL.md or plugin.json.
  4. Invoke /continua:analyze (namespace = plugin.json → name).
  5. Read the current usage guide with learn_how_to_use and check the connection through /mcp.

Validate before submit

From the plugin root:

claude plugin validate .

Resolve any reported issues before submission. See the manifest reference for supported fields and paths.


Submit to the community marketplace

Follow the current plugin publishing and submission guide. The directory submission route is the developer portal:

  1. Validate the plugin from the repository root.
  2. Submit the GitHub repository through the portal and complete its validation and review steps.
  3. After publication, share the install instructions from the published listing.

analyze skill

skills/analyze/SKILL.md teaches Claude when and how to use Continua MCP through three main sections:

  • Investigation: Review very large source sets with up to 1 billion tokens of working context, as described in the Platform overview. Reconcile financials, agreements, amendments, invoices, and other evidence; surface conflicts and missing support with citations. Produce Word memos, Excel workbooks, PowerPoint presentations, and JSON/CSV outputs. Use Agent mode for project-aware orchestration.
  • Project Room: Upload related materials once and reuse the same context for reports, follow-ups, and new questions. Share the source set with colleagues and other AI clients through project access settings. Preserve reviewed decisions, definitions, policies, recipes, and expert corrections in an Enterprise Second Brain, a shared LLM wiki that people and AI can read, refine, and reuse across investigations. Users can also forward requests and case attachments to query@agent.askcontinua.com through Continua Inbox.
  • Automation: Schedule credit memos and supply chain reviews against project sources and connected systems. Follow Intake → Underwrite → Monitor → Policy: review changes and exceptions, confirm or correct conclusions, and preserve approved lessons for the next case.

Requires Continua MCP connected; the plugin loads the connection through .mcp.json. Financial conclusions and policy changes remain subject to the responsible team's review.

Industry workflows

WorkflowSource materialsDecision-ready output
Private creditBorrower financials, debt schedules, credit agreements, amendments, compliance certificates, prior memosCited credit memo, effective covenant and headroom matrix, cash-flow reconciliation, changes since the last review, and open diligence items
Factoring and receivables financeReceivables aging, invoices, credit notes, remittances, purchase orders, delivery evidence, customer agreementsReceivables reconciliation, potential eligibility exceptions, dilution and concentration analysis, disputed items, and a borrowing-base review using the supplied facility rules
Supply chain and working capitalSupplier/customer contracts, orders, invoices, payment records, inventory and delivery reportsSupplier and buyer exposure map, payment and delivery exceptions, contract obligations, working-capital risks, and a prioritized follow-up plan
Enterprise knowledge managementApproved policies, metric definitions, prior decisions, SOPs, templates, expert reviewsReusable project knowledge, cited decision history, policy comparisons, proposed lessons for review, and a consistent starting point for the next case

Example requests

/continua:analyze Review our private credit project against the current loan
agreement and all amendments. Reconcile reported covenant compliance, identify
changes since the prior review, and produce a cited credit memo plus an Excel
exception register. Apply our approved credit policy and flag missing evidence.

Other starting points:

  • “Reconcile this factoring pool's invoices, collections, credit notes, and aging. Apply the supplied eligibility rules and show every unresolved exception.”
  • “Review supplier exposure and working-capital risks across contracts, orders, invoices, and payment records. Prepare a management deck and action register.”
  • “Turn this reviewed case and my corrections into a draft decision record and proposed project knowledge. Show the changes for review before saving an approved policy.”
  • “Run a weekly credit review against this Project and compare each result with the prior review.”

Other clients (Cursor, Codex, ChatGPT)

This repository is a Claude Code plugin. For other clients, connect the Continua MCP server using the MCP connection guide.

Continua AI · MCP tools and connection guide · Platform · Enterprise Second Brain and LLM wiki · Privacy policy

Solutions & use cases: All workflows · Finance underwriting · Credit review · Multi-document processing · Knowledge graph.