Claude Code
Retrieve project requirements, decisions, and source evidence during planning and implementation.
Integrations
Connect project context once. Use it across your coding tools.
Developers and AI agents can retrieve the same requirements, architecture decisions, documentation, and project memory through MCP, REST API, CLI, and the Web portal.
Each client or application can use the existing retrieval layer instead of building and maintaining its own.
Coding tools
One ContextHarbor project. Every approved tool.
Give Claude Code, Codex, Cursor, Cline, OpenCode, and custom agents access to the same project-scoped context without locking knowledge inside one IDE or model.
Codex
Bring architecture rationale and project memory into repository work across sessions.
Cursor
Use the same governed project context from an IDE-centered development workflow.
Cline
Connect agentic coding work to knowledge that lives beyond the repository.
OpenCode
Keep an open coding workflow grounded in the same project sources and decisions.
Custom agents
Use Streamable HTTP MCP or the REST API for internal engineering applications and approved automation.
Available interfaces
Choose the access pattern that fits the workflow.
Remote MCP
Connect compatible team AI clients over HTTPS with an MCP key that respects project access.
REST API
Add project-scoped search and optional AI Answer to internal applications, portals, automations, and custom AI services.
Web portal
Manage users, projects, documents, live indexing, search, MCP setup, and observability from the browser.
GitHub connectors
Synchronize a selected repository, branch, and optional path into the standard project ingestion and retrieval pipeline.
Agent skills
Install focused retrieval guidance that helps compatible AI assistants query, scope, and inspect ContextHarbor evidence effectively.
Administrator-approved providers
Connect named embedding and OpenAI-compatible LLM providers only after an administrator enables and configures them.
Connection flow
Connect a client without rebuilding the retrieval layer.
Use the interface and credential model that fit the selected team workflow.
Select project
Choose the knowledge boundary the client or application should search.
Grant project access
Use an MCP API key or application credential with access to the required project.
Connect client
Configure Remote MCP, call the REST API, or work through the Web portal.
Retrieve source-backed context
Return relevant project passages with source information for inspection.
Integration principles
Keep every connection intentional.
Defined interfaces
Connect clients and applications through the interface selected for the workflow.
Explicit project scope
Select the relevant project rather than querying the full knowledge estate.
Source preservation
Keep retrieved material connected to the source available for inspection.
Keep credentials server-side
Keep integration credentials outside browser code and apply the network controls required by your deployment.
Third-party compatibility depends on the interfaces and authentication methods supported by the selected client.
MCP tools
One MCP server. Tool groups that switch on with the deployment.
The server always advertises the core retrieval set and adds more groups when the deployment enables them — a full-tenant deployment exposes around forty tools in all.
Tool groupWhat it adds
Core retrievalIngest,
query_documents, search_project_docs, read_chunk_neighbors, list_projects, get_project_brief, planning_context, requirement_lookup, status.Knowledge hierarchy
get_project_structure, get_area_brief, get_knowledge_item, search_area for consolidated, source-cited briefs. Knowledge Briefs.Durable memorySession lifecycle, verified handoffs, recall, and provenance checks. Memories.
Prompt libraryList, fetch, and manage a project's governed prompts.
Code intelligence
setup_code_graph, code_context, code_impact, get_code_brief. Code Intelligence.ConnectorsGitHub repository sync — configure a personal access token, pick a branch and optional path prefix, and sync files into the project pipeline.
AdminSettings, MCP keys, observability, storage, and client-usage endpoints.
GitHub connectors
Sync a repository into the same project pipeline.
Add a connector with a personal access token, pick a branch and optional path prefix, and Sync Now — synced files land in Files and are chunked, embedded, and searchable like uploads.
ContextHarbor
Discuss your team’s knowledge workflow.
See how ContextHarbor could support a controlled operational or engineering knowledge workflow within your environment.
Pricing depends on deployment, integration, and support requirements.