Skip to content

Connect Asana to Claude for a project team

How to connect Asana to Claude so each person sees only their own projects, every tool that deletes is blocked, and every call is recorded.

Pranasri Marripudi 7 min read
A project operations lead reviewing Asana task access for a team in an AI assistant

The situation

A project team already lives in Asana. Someone on that team asks Claude to summarize what slipped this week, and now you have to decide what Claude may touch. The useful version lets a coordinator create a follow-up task and add a comment. The version that gets you called into a meeting lets anyone delete a project or remove a teammate from a workspace.

This playbook is how a project operations lead decides to connect Asana to Claude for a whole team. The order matters: identity first, then what the role may reach, then the record of what happened. MCP is the open protocol that lets an AI client call tools in other systems. Elaichi is where you decide which of those tools each person reaches.

How do you connect Asana to Claude for a whole team?

You connect Asana to Claude through Elaichi's MCP endpoint, and each person connects their own Asana account. Claude never holds an Asana credential. It holds an Elaichi token, and Elaichi holds the connection.

The Asana connector is one Elaichi authors and runs itself, not a vendor's own MCP server. It carries 190 tools and signs in with OAuth or an API key. Use OAuth here, because OAuth signs in as a person. Every call through a connection uses the account that connection was made with. An OAuth connection made by a coordinator acts as that coordinator in Asana.

The steps, in the order you do them:

  1. An Org Owner or Org Admin opens the Asana connector in Elaichi and confirms sign-in works.
  2. Each team member adds their own Asana connection, keeping it private rather than sharing it.
  3. Each person connects Claude to Elaichi once and approves their own grant on Elaichi's consent screen.
  4. You write one restriction on the project role, covered below.
  5. You check the audit log after the first day of real use.

The MCP endpoint is https://api.elaichi.ai/mcp, the same address for every organization. Claude, ChatGPT, Cursor and Microsoft Copilot all sign into it. There is no per-team or per-person URL to hand out, and nothing to paste into a config file. The MCP specification sets out how a client obtains an access token before it calls a server (checked November 2026, the MCP authorization specification).

Why does each person connect their own Asana account?

Because a connection runs on the account behind it, not on the account of whoever asks. If Tara connects Asana with her own sign-in, her Claude reaches the projects Tara can see in Asana. If she instead used a connection shared by someone with broader reach, her Claude would act as that other person.

That is the whole mechanic, and it decides your rollout. For tools that run on each person's own account, each person needs their own connection. A toolbox shared with a team runs its entries on the connection pinned in them, under the authority of whoever pinned it. That is useful when you want one fixed account, and wrong when you want per-person visibility. The same pattern holds for Notion pages each person can already see.

In practice this means telling the team not to share Asana connections with each other. A private connection reaches only its owner. When they connect Claude, each person approves a grant that covers their own connections. Claude's reach then tracks their Asana reach, and you maintain no second map of who sees what.

Asana's OAuth documentation covers how an application signs a user in and then acts with that user's authorization (checked November 2026, Asana's OAuth guide). Asana's permissions overview describes how membership of a team or a project decides what a person can see there (checked November 2026, Asana's permissions overview).

Which Asana tools should you block?

Block every tool that deletes, and leave the ones that create and update tasks alone. A project team needs to add a task and comment on it. It does not need to delete a portfolio from a chat window.

Write one block restriction on the project team's role. Add the Asana connector, choose Specific tools, and use the Block deleting preset. The preset blocks every tool on the connector that deletes, rather than a list you maintain by hand. The Asana connector carries 33 delete and remove tools, so a handful of names would leave most of them open.

The tools the preset covers include these:

  • delete_a_asana_task_by_id
  • delete_a_asana_project_by_id
  • delete_a_asana_section_by_id
  • delete_a_asana_portfolio_by_id
  • delete_a_asana_goal_by_id

The two member-removal tools, asana_workspaces_remove_user and asana_teams_remove_user, remove people rather than records. Tick both by name in the same rule, so the rule covers them whether or not the preset counts a removal as a delete.

Leave create_a_asana_task, update_a_asana_task_by_id and create_a_asana_story reachable, so the assistant can file follow-ups and add comments. Reads such as list_all_asana_workspaces, list_all_asana_users and list_all_asana_tags stay reachable too. The Block deleting preset leaves reading on.

Two mechanics decide how the rule behaves. First, in Elaichi, restriction targets are role or user only; the organization default is the absence of a rule, which means allow everything. A rule aimed at a team's members is written on the role they hold, or on each member. Second, a layer with only block rules has no allowlist, so everything it does not block passes it. A block list holds only the tools it names, so an Asana tool the rule does not name stays reachable, and a new tool added later would need adding.

If you would rather hold the role to a named set instead, you can write allow rules. Be careful: once a role carries any allow rule, every connector and tool none of its allow rules names is blocked for that role. A role held to Asana reads by an allow rule also needs allow rules for every other app it uses. The trade-off is strictness against maintenance, and most project teams are better served by the block list.

Restriction changes take effect within about two minutes. Write the rule before you tell the team to start.

What does Claude see, and what does it not?

Claude sees Elaichi's search and run tools, not a list of 190 Asana tool names. In Elaichi, connected tools are never listed one by one, however few there are. The model finds a tool with search_tools and runs it with read_tool for a read or execute_tool for anything else.

A restricted tool is not invisible. search_tools names a restricted match and flags it as restricted, with no schema and no way to call it. The result also carries a count of how many matches were withheld. Calling a blocked tool by name returns a refusal, not a confusing "unknown tool". The refusal says the tool is restricted for the person's role, with a way to ask an admin for access. That is better for a coordinator who needs to understand why Claude declined.

One honest limit: Elaichi's MCP endpoint has no prompt-injection defense of its own. Elaichi reads a tool name and arguments from a call, never the person's prompt. It cannot judge a call against what the person actually asked for. That defense belongs to the client. The controls Elaichi gives you here are the grant's scopes, the role's restrictions, and the record afterwards.

What lands in the audit log?

Every connected-tool call that reaches execution is written to Elaichi's audit log. The row records the person, the tool's advertised and catalog names, the connector, and the connection the call reached. It also records the tool's classification, the surface the call came from, and the MCP client's id when the surface is mcp. A failed call records an error code.

The console's detail panel reads as a sentence: who did what to which item through which connection. It opens with where the call ran, such as "Using Claude,". That is the line you paste into a ticket when someone asks what the assistant did on Tuesday.

Two things to know before you rely on it. A call refused at the scope check writes no entry, because that refusal happens before any executor runs. The log is a record of what ran, not of everything attempted. And Elaichi keeps 90 days of audit history. If your retention policy needs longer, forward it to your own log service. Logging destinations come with the Black plan, and Datadog is the destination that delivers. For what a log entry should contain in the first place, see what an AI agent audit log must capture.

When Asana's own tools are enough

If Asana is the only app your team wants in Claude, you may not need Elaichi at all. That holds when everyone on the team is comfortable connecting it themselves. Asana also documents an MCP server of its own (checked November 2026, Asana's MCP server docs). Connecting one app to one assistant is a smaller job than running a control plane, and a smaller job deserves a smaller answer.

The case for one address starts when the second and third apps arrive. Then you are maintaining separate approvals per app, separate per-app logs, and a separate offboarding step for each. Elaichi's two edges are one address that Claude, ChatGPT, Cursor and Microsoft Copilot all sign into, and one governed catalog behind it. One audit log covers every connector, which is what makes a developer's Jira, GitHub and Slack together one setup rather than three.

The other case is the tool-level block. If the only control you have is on or off for a whole app, you cannot keep create_a_asana_task open while delete_a_asana_project_by_id stays shut. If the whole app on or off is fine for your team, take the simpler path. For the wider question, read when you don't need an MCP gateway yet.

When someone leaves the team

Removing a member from the organization revokes every live MCP grant they hold there, inside the same call that deletes the membership. Because grants are re-read on every request, their Claude fails on the next call rather than after a cache window. Reactivating them later does not bring the grants back; they consent again.

Their private Asana connections are a separate decision. Elaichi's offboarding preflight lists what the departing person owns before anything changes, including connections and the toolbox entries elsewhere that rely on them. A private connection nothing else depends on is deleted with the member by default. Removing someone from Elaichi does not remove their Asana account, which is deprovisioned in Asana or through your identity provider. Offboarding when the agent holds access walks the rest of it.

To see which connectors sit beside Asana for the rest of the company, browse the connector catalog.

FAQ

Frequently asked questions

Does connecting Asana to Claude through Elaichi let Claude see projects a person cannot see in Asana?

No. In Elaichi, each person connects their own Asana account with OAuth, and every call through a connection runs on the account that connection was made with. Claude reaches only what that person's Asana account can reach. If instead someone uses a connection shared by a colleague, the calls run as that colleague's Asana account, which is why per-person visibility requires per-person connections.

Can you block Asana deletes while still letting an AI assistant create tasks?

Yes. In Elaichi, a restriction rule written on a role can name individual tools of a connector, and the Block deleting preset blocks every tool on that connector that deletes. Add the two member-removal tools, asana_workspaces_remove_user and asana_teams_remove_user, by name in the same rule. Reading stays on with that preset, and create_a_asana_task and update_a_asana_task_by_id stay reachable. A layer with only block rules blocks nothing it does not name.

How long does a restriction change take to apply in Elaichi?

A restriction change reaches every surface within about two minutes. Role, team-membership and access-grant changes run on the same schedule, so write a rule before telling a team to start rather than during their first session. Some changes are faster: removing or suspending a member, revoking a share, and disconnecting an account all take effect on the next call.

Is the Elaichi Asana connector a vendor MCP server?

No. The Asana connector is one Elaichi authors and runs itself. Elaichi's catalog holds two kinds of connector: ones Elaichi writes and operates, which are most of the catalog, and native MCP connectors, where the vendor builds and runs its own MCP server and Elaichi handles sign-in, access and audit in front of it. Asana is the first kind. It carries 190 tools and signs in with OAuth or an API key.

What does Elaichi record when Claude calls an Asana tool?

Every connected-tool call that reaches execution writes an audit row naming the person, the tool, the connector, the connection the call reached, the tool's classification, the surface the call came from, and the MCP client's id for calls over MCP, plus the outcome and an error code on failure. A call refused at the scope check writes no row, because that refusal happens before any executor runs. Elaichi keeps 90 days of audit history.

Put AI to work in your own apps

14 days of Black, no credit card. Start with one app and one team.

When the trial ends, nothing is deleted. You move to the free Personal plan, and connections, roles and the audit log stay where they are.

Add one link to your AI app

The same link for everyone on your team.

https://api.elaichi.ai/mcp
  • ChatGPT
  • Claude
  • Cursor
  • Microsoft Copilot