Which issue trackers don't charge for AI agents? A meter audit you can run today

Three issue trackers say in their docs or code that AI agents are not billed as seats: Radial, Linear, and Jira when the agent runs as a service account (5 are free; more need Atlassian Guard). GitHub does not list GitHub Apps among its billed users, so an agent that runs as a GitHub App is not a seat either. Seat rules are half the bill, though. Linear's own agent work and Jira's enriched MCP searches draw usage credits, and Asana, ClickUp and Plane all ship AI credit pools. Of the trackers below, Radial is the only one that both keeps agents off the seat count and has no usage meter: $50 per user per year, flat, with nothing to self-host. It is not the cheapest. GitHub Team is $48 per user per year, and Linear Free costs $0 for a small team under 250 issues.

That answers the question for today's price pages. The rest of this post is the part you can reuse when the pages change: a meter audit that takes about an hour, tells you which line of your tooling bill scales with agent use, and ends with a number you can put in a budget doc.

#What a meter is, and why agents make it matter

A meter is any line on an invoice that moves with use instead of with headcount: AI credits, tokens, requests, per-run fees, or a seat that an agent account quietly occupies. A team of 10 people running 20 agents has the same 10 seats in January and December. Its agents do not stay constant. They file more issues, run more searches and trigger more automations every month you give them more work.

So the question to ask a vendor has two parts. Does an agent with its own account count as a seat? And does an agent's call draw from a credit pool that bills past a limit? A "no" to both means the bill tracks your people. A "yes" to either means the bill tracks your agents, and you need a cap before you need a bigger budget.

#How trackers and dev tools bill AI agents and usage

Every row below comes from the vendor's official pricing page or docs, fetched on Nov 3, 2026, unless the row says Nov 2 (those figures are reused from what is a cost-effective issue tracker, which sourced them a day earlier). Prices are list prices in USD. Where a vendor does not publish a rule, the cell says "not published".

Two things the table does not show. First, most credits are drawn by the vendor's own AI features, not by your agents. Claude Code or Codex calling Linear's API or MCP server draws no Linear credits, per the pages we read. Jira is the exception that matters: an external agent that calls the Rovo MCP search tool spends Rovo credits on every call, up to 10 at a time. Second, "not published" is not "no". If you plan to give each agent its own account on Asana, ClickUp, Plane or Shortcut, ask the vendor in writing before you count on a $0 agent.

For the full seat math at 1, 10 and 25 people, see what is a cost-effective issue tracker. For Jira's add-ons and admin time, see the real cost of Jira.

#The meter audit, step by step

The audit has five steps. The first takes under five minutes and needs no account anywhere. You end with a filled-in worksheet and a decision.

#Step 1: list every tool your agents touch (5 minutes)

Your agents tell you which tools they reach. Start with the agent's own config and your git history:

# MCP servers your Claude Code agent loads
claude mcp list

# Project-level MCP config, if the repo has one
cat .mcp.json 2>/dev/null

# Bot and agent identities that committed in the last 90 days
git log --since=90.days --format='%an <%ae>' | sort | uniq -c | sort -rn | head -20

The last command is the one people skip. Every [bot] address, service account and shared "ci@" user in that list is an identity some vendor may bill. Write each tool on a row: tracker, code host, coding assistant, CI, docs, chat. Include anything an agent reads from or writes to, even once a day.

#Step 2: find the seat rule and the meter for each tool (20 minutes)

For each row, open the vendor's pricing page and billing docs and answer four questions. Copy the vendor's exact sentence and the URL next to each answer, the way the table above does.

1. Does an agent with its own account count as a seat? Search the docs for "agent", "bot", "service account", "app user" and "billable". If the answer is not published, write "not published", not "no".
2. Is there an AI credit pool? Note the allowance per user or per workspace, whether it is pooled, and whether it resets monthly.
3. What does usage past the pool cost? Write the unit price ($0.01 per Jira Rovo credit, $0.60 per Asana on-demand request, $10 per 10,000 ClickUp credits).
4. Is overage on by default? Jira's is. That one fact decides whether a busy month bills you or stops your agents.

#Step 3: price N humans and M agents, then double it (15 minutes)

Fill one row per tool. Count the agents that would need a seat if the vendor billed them (column E is 1 for "yes" or "not published", 0 for a documented "no"). For credits, estimate the agents' monthly use from real numbers: calls per agent per workday, times 21 workdays, times credits per call.

Seats cost B x (C + D x E). The meter costs (G minus F) x H when G is larger than F, and nothing otherwise. Then double G, because agent use grows, and see what moves. Save the worksheet as meter.csv and let awk do the arithmetic:

cat > meter.csv <<'EOF'
tool,seat_usd_month,humans,agents,agents_are_seats,credits_included,credits_used,usd_per_credit
jira-standard,9.70,10,20,0,250,21000,0.01
linear-basic,10,10,20,0,0,0,0
github-team,4,10,20,0,0,0,0
radial,4.1667,10,20,0,0,0,0
asana-starter,10.99,10,20,1,0,0,0
EOF

awk -F, 'NR>1 {
  seats = $2 * ($3 + $4 * $5)
  over1 = ($7     > $6) ? ($7     - $6) * $8 : 0
  over2 = ($7 * 2 > $6) ? ($7 * 2 - $6) * $8 : 0
  printf "%-14s seats $%7.2f/mo  meter $%7.2f/mo  meter at 2x $%7.2f/mo\n", $1, seats, over1, over2
}' meter.csv

The output for the example rows:

jira-standard  seats $  97.00/mo  meter $ 207.50/mo  meter at 2x $ 417.50/mo
linear-basic   seats $ 100.00/mo  meter $   0.00/mo  meter at 2x $   0.00/mo
github-team    seats $  40.00/mo  meter $   0.00/mo  meter at 2x $   0.00/mo
radial         seats $  41.67/mo  meter $   0.00/mo  meter at 2x $   0.00/mo
asana-starter  seats $ 329.70/mo  meter $   0.00/mo  meter at 2x $   0.00/mo

Read it the way finance will. In this scenario Jira's seats are $97 a month and its meter could reach $207.50, then $417.50 when agent searches double. Most months will be lower: 10 credits is the documented ceiling per search, and writes and single-issue lookups draw none. The point is that the number can move, and only a cap decides how far. The Asana row shows the other risk: if agents with their own accounts turn out to be seats, 10 people cost the same as 30. Both scenarios are ours, not vendor quotes; the inputs are in the worksheet so you can replace them with your own.

#Step 4: set alerts and caps before the next billing cycle (15 minutes)

For every tool with a meter, turn on the alert and set a cap at the number from step 3. Where the vendor lets you, choose "stop" over "keep billing" at the cap, and decide in advance who gets the alert.

• Jira (Rovo credits). In Atlassian Administration, set a spending limit on extra usage, or turn extra usage off so billable interactions pause at the allowance. Org and billing admins get notices at 80% and 100%. To see what your agents already spend, go to Usage, then Meters, select Rovo credits and export the CSV (up to 3 months of history).
• Linear (AI credits). In Settings, Usage & limits, set a spend limit. If you use Automatic Reload, its threshold is your real cap: the minimum auto top-up is $50. Failed and retried runs are billable for the resources used.
• GitHub Copilot. Set a budget for usage-based billing. On Business and Enterprise, decide whether paid usage past the included credits is allowed at all.
• Asana. Set the spending cap on on-demand billing; alerts fire at 65%, 85% and 100%. You can also turn AI requests off per user.
• ClickUp. Check your plan's daily MCP call cap against step 3. At 300 calls a day on Core, 20 agents get 15 calls each before the cap stops them.

Then put the date in your calendar for the week after the first bill and compare the real meter line with your estimate.

#Step 5: decide (10 minutes)

Write the result next to the seat price in your budget doc and pick one of three outcomes:

1. Keep the tool. The meter is small, capped and alerting. This is the right answer for most teams, most of the time.
2. Change how agents connect. Move agents to a person's token, a service account or a GitHub App where the vendor documents those as non-billable, and keep agents off the vendor's metered features (on Jira, avoid cross-product search in agent loops).
3. Move the agent work to a tool with no meter. Do this only when the meter line beats the switching cost. Count the migration day, too: migrating from Linear and migrating off Jira list what comes across and what does not.

Run the audit again each quarter, and any time a vendor emails you about "usage-based pricing". Jira's overage start date, December 3, 2026, is the next one on the calendar.

#Where Radial wins, and where it loses

Radial's bill is one number: $50 per human per year, billed annually. Agent members never count as seats, and there are no credits, tokens or AI requests to meter, because Radial runs no AI of its own. Your agents connect through the radial CLI, the hosted MCP server at mcp.radial.build or the REST API, and the invoice is the same whether they make ten calls a day or ten thousand. The commitment is written into the Plain Software Pledge: if Radial ever ships a copilot, meters your usage or charges you for AI you didn't ask for, we refund your subscription in full (manifesto).

It is the wrong pick in these cases, and the meter audit will not change that:

• GitHub Team is cheaper. $48 per user per year against Radial's $50, with agents as GitHub Apps off the seat count. If your work lives in GitHub and a flat issue list is enough, stay.
• Free tiers beat any paid plan for small teams. Linear Free has unlimited members up to 250 issues, Jira Free covers 10 users, and Shortcut Free covers 10. Radial has no free tier and no trial.
• Linear also keeps agents off the seat count, and your external agents draw no Linear credits. If you do not use Linear's own coding sessions or Loops, Linear has no meter for you either; the difference is the price ($120 per user per year on Basic) and the missing first-party issue CLI. See the best Linear alternative.
• Radial bills a year upfront. There is no monthly plan, adding a person mid-term is invoiced at once, and removing one gives no credit.
• Feature gaps. No SAML or SCIM, no webhooks you can register through the API, CLI or MCP, and agents poll for work rather than receive a push.

#FAQ

#Which issue trackers do not charge per seat or usage credits for AI agents?

As of Nov 3, 2026, Radial is the one tracker we found that both keeps agents off the seat count and has no usage credits: agent members never count as seats, and nothing is metered, at $50 per user per year. Linear does not bill agents as seats and your external agents draw no Linear credits, but Linear's own coding sessions and Loops draw prepaid AI credits. Jira does not bill service accounts (5 free, more with Atlassian Guard), but an agent's Rovo MCP search can spend up to 10 Rovo credits, at $0.01 per credit past the pool from December 3, 2026. GitHub Issues has no meter and does not bill GitHub Apps as users.

#Which issue tracker doesn't charge for AI agents as seats?

Radial, Linear and Jira say so in their code or docs. Radial excludes agent members from the seat count. Linear's docs say "Agents are not counted as billable seats in Linear." Atlassian says service accounts "do not count towards your user limit", with 5 free and up to 250 on Atlassian Guard Standard at $4.35 per user per month. GitHub does not list GitHub Apps among billed users. Asana, ClickUp, Plane and Shortcut do not publish a rule for agents with their own accounts.

#How do I find out whether my tools bill AI agents as seats?

Run git log --since=90.days --format='%an <%ae>' | sort | uniq -c | sort -rn to list the bot and agent identities in your repo, and claude mcp list to list the tools your agent loads. Then search each vendor's billing docs for "agent", "bot", "service account" and "billable", and copy the exact sentence. If the docs say nothing, treat the agent as a seat in your budget until the vendor confirms otherwise in writing.

#How much can AI credits add to a tracker bill for 10 people and 20 agents?

It depends on the tracker and on whose AI the agents use. In our worked scenario on Jira Standard, 20 agents each running 5 Rovo MCP searches a workday at the documented ceiling of 10 credits could spend 21,000 credits a month: $207.50 past the 250-credit pool, against $97 a month in seats. On Linear, GitHub Issues and Radial, external agents draw no tracker credits, so the meter line is $0.

#How do I stop AI credit overages in Jira?

In Atlassian Administration, set a spending limit on extra usage or turn extra usage off, so billable interactions pause when the monthly allowance runs out. Extra usage is on by default from December 3, 2026. Admins get notices at 80% and 100% of the allowance, and Usage, then Meters, then Rovo credits exports a CSV of what you already spend.

#One number to put in the budget doc

Run step 1 now: git log and claude mcp list take five minutes and need no signup. If the audit ends with a meter line you would rather not manage, Radial's pricing is one number, $50 per user per year, and it is the same in December as in January.