Platform / Agent Harness

Use the same agent assets through BotDojo, CLI, and MCP.

Agent Harness is how BotDojo keeps your context, tools, and permissions reusable across native agents, command-line workflows, and MCP-compatible environments.

Example CLI flow
botdojo-tool register bdres_... --api-url https://api.botdojo.com
botdojo-tool search "support playbook"
botdojo-tool open "mcp://workspace"
botdojo-tool call "mcp://workspace" read_file '{"path":"README.md"}'

The same underlying assets can also be mounted into BotDojo agents or surfaced through MCP-compatible clients.

Harness Modes

One asset layer, multiple ways to use it

BotDojo agents

Agents can mount model contexts, discover resources mid-run, and enable the tools they need without hard-wiring every capability into the base prompt.

CLI access

The botdojo-tool CLI can register a resource token, search, read, open, and call tools directly while reusing refresh and re-approval behavior.

MCP delivery

Model contexts expose MCP-style cards, prompts, resources, and tools so the same package can be consumed by BotDojo and MCP-compatible clients.

Existing workflows

Teams can keep working through familiar environments and workflows around tools like Claude Code, ChatGPT, or internal assistants instead of switching everything at once.

Why It Matters

Your team should not have to rebuild the stack for every surface

Agent Harness reduces the friction of moving between BotDojo, CLI workflows, and external clients because the transport changes without forcing the asset layer to change with it.

The same resource URIs and model context cards travel across surfaces
Prompts, tools, documents, files, and integrations do not need a separate copy per client
Approval and permission behavior stays tied to the same underlying resource model
Agent Harness - BotDojo