featureloom MCP tools
What the agent calls, and with which right.
featureloom runs a remote MCP server with 27 tools. Every console, Claude Code, Codex, Cursor or another MCP client on your machine, gets the reading tools, and the rest by the rights of the person it acts for. The server refuses any call outside those rights.
Address: https://mcp.featureloom.com/mcp. A console sends its credential in the Authorization: Bearer header.
Server and connection
One address, three ways to connect.
A person’s console or an automation
Every console connects to the same address and sends its credential in the Authorization: Bearer header. A person connects a console with a connect command from the app, holding an flc_ secret that works for 90 days and expires sooner after 14 days without use, or signs it in in the browser over OAuth 2.1. That sign-in returns only to the console on the person’s own machine. An automation, for example in CI, has its own fla_ key.
- For Claude Code and Codex the app gives you a ready connect command, and for Cursor a button and an entry for its file
- Another MCP client that runs on your machine signs in the same way
- Each person has up to 10 consoles at once in an organization
# Claude Code, then /mcp inside it
claude mcp add --scope user --transport http featureloom \
https://mcp.featureloom.com/mcp
# Codex
codex mcp add featureloom \
--url https://mcp.featureloom.com/mcp
# Cursor CLI, with the entry in ~/.cursor/mcp.json
agent mcp login featureloomTools by right
Every console reads. The rights give the rest.
The server has 27 tools, and each belongs to one group: reading or one of four rights. The rights carry the names an administrator sees on the People screen. A console’s tool list follows its person’s rights and refreshes when the console reconnects.
Reading · 10 tools, every console and every automation
| Tool | What it does |
|---|---|
list_plans | The organization’s plans without their documents, with dependencies and readiness; filter by project, state, column or set aside. |
get_plan | One plan whole: its document, dependencies and acceptance verdict. |
list_projects | The organization’s projects with their plan counts. |
list_epics | The epics a project’s plans are grouped into, with their plan counts. |
get_plan_template | The project’s plan template, the shape your team writes plans in. |
get_acceptance | A plan’s latest acceptance verdict, and whether a rejection still waits for rework. |
list_prompts | The organization’s prompts: its process, roles and rules. |
get_prompt | One prompt whole; with a plan key, also the whole plan and the documents of its delivered predecessors. |
next_plan | The next ready plan on the board, in the order people see. |
whoami | Who the console or automation is right now: the person, the organization, the rights and the tool list. |
list_plans
The organization’s plans without their documents, with dependencies and readiness; filter by project, state, column or set aside.
get_plan
One plan whole: its document, dependencies and acceptance verdict.
list_projects
The organization’s projects with their plan counts.
list_epics
The epics a project’s plans are grouped into, with their plan counts.
get_plan_template
The project’s plan template, the shape your team writes plans in.
get_acceptance
A plan’s latest acceptance verdict, and whether a rejection still waits for rework.
list_prompts
The organization’s prompts: its process, roles and rules.
get_prompt
One prompt whole; with a plan key, also the whole plan and the documents of its delivered predecessors.
next_plan
The next ready plan on the board, in the order people see.
whoami
Who the console or automation is right now: the person, the organization, the rights and the tool list.
Realizes plans · 5 tools, a new person has it from the start
| Tool | What it does |
|---|---|
start_plan | Takes a ready plan on: moves it from the column marked backlog to the one marked in realization and records who works on it. |
report_realized | Reports a plan delivered, with or without a note. |
rework | Takes a rejected plan back into realization. |
add_note | Adds a progress note to a plan in realization. |
ask_question | Asks people a question about the plan in realization; the question goes into the plan document. |
start_plan
Takes a ready plan on: moves it from the column marked backlog to the one marked in realization and records who works on it.
report_realized
Reports a plan delivered, with or without a note.
rework
Takes a rejected plan back into realization.
add_note
Adds a progress note to a plan in realization.
ask_question
Asks people a question about the plan in realization; the question goes into the plan document.
Writes plans · 9 tools
| Tool | What it does |
|---|---|
create_plan | Creates a plan as a document; it lands in the column marked backlog. |
create_epic | Creates an epic that groups a project’s plans. |
add_link | Links two plans as a dependency or a relation. |
remove_link | Removes one dependency or one relation between plans. |
update_plan | Changes a plan’s document, name or type. |
archive_plan | Moves a delivered plan to the project’s archive. |
restore_plan | Brings a plan back from the archive to the column of delivered plans. |
shelve_plan | Sets a plan aside, off the board and out of next_plan. |
unshelve_plan | Brings a plan that was set aside back to the board. |
create_plan
Creates a plan as a document; it lands in the column marked backlog.
create_epic
Creates an epic that groups a project’s plans.
add_link
Links two plans as a dependency or a relation.
remove_link
Removes one dependency or one relation between plans.
update_plan
Changes a plan’s document, name or type.
archive_plan
Moves a delivered plan to the project’s archive.
restore_plan
Brings a plan back from the archive to the column of delivered plans.
shelve_plan
Sets a plan aside, off the board and out of next_plan.
unshelve_plan
Brings a plan that was set aside back to the board.
Edits prompts · 1 tool, only an administrator’s console changes a command name
| Tool | What it does |
|---|---|
save_prompt | Creates a prompt or saves a new version of it. |
save_prompt
Creates a prompt or saves a new version of it.
Deletes permanently · 2 tools, in a console only together with Administrator
| Tool | What it does |
|---|---|
delete_plan | Permanently deletes a plan still in the column marked backlog, with its versions. |
delete_prompt | Permanently deletes a prompt with its versions. |
delete_plan
Permanently deletes a plan still in the column marked backlog, with its versions.
delete_prompt
Permanently deletes a prompt with its versions.
How rights work
An administrator grants the rights. The server enforces them.
Five rights for a person, four for their consoles
A person has five rights: Administrator, Realizes plans, Writes plans, Edits prompts and Deletes permanently. Their consoles act with the last four, and an automation has its own four. A new person has Realizes plans from the start; an administrator grants the others on the People screen.
- A change of rights applies from the next call, and a console refreshes its tool list when it reconnects
- Only a person gives the acceptance verdict, in the app; the server refuses one sent from a console
- The agent chooses neither the column nor the position: starting and reporting move the card only between the columns marked with a state
Questions about the MCP server
Can I connect another MCP client?
Yes, if it runs on your machine and supports OAuth sign-in. Add the server at https://mcp.featureloom.com/mcp in it and sign in in your browser. For Claude Code and Codex the app gives you a ready connect command, and for Cursor a button and an entry for its file.
What happens when I change a person’s rights?
The server applies the new rights from the next call. The console’s tool list refreshes only when it reconnects, and until then the server refuses any call outside the rights. A change of rights disconnects no console.
How do I connect an automation, for example in CI?
An automation is the organization’s own account for unattended jobs, with its own rights and keys with a term. An administrator creates it on the Automations card in Integrations, and the job sends its fla_ key in the Authorization: Bearer header.
Can an agent accept its own work?
No. An AI agent delivers and reports the work, but only a person gives the acceptance verdict, in the app, and the server refuses one sent from a console or an automation. The agent reads the verdict with get_acceptance and, after a rejection, takes the plan back with rework.
Give your agents the tools. Keep the rights.
30 days free, no payment card needed. You connect a console in a few minutes.