Plans & query quotas
Community, Pro, Pro+, and Teams: what the graph-query quota actually meters, how the meter is enforced, and why multi-device use is never a problem.
Memtrace meters one thing: graph-query calls against the MCP tool surface — search, impact analysis, history, quality, review, architecture. Indexing, watching, fleet coordination, and diagnostics are never billed, on any plan. Everything runs locally; the meter counts calls, it never ships your queries anywhere.
Plans#
| Plan | Price | Graph queries | Notes |
|---|---|---|---|
Community | Free | 1,000/month | Then a server-set daily allowance (currently 50/day) instead of a hard stop. |
Pro | $9 | 5,000/month | Same billable tool list as Community, higher ceiling. |
Pro+ | $15 | 20,000/month | For heavier daily agent use. |
Teams | $29/seat | Unlimited | Never quota-gated, plus self-hosted MemDB — see Deploy MemDB. |
Plan ids in the product are community, pro, pro_plus, and teams — there is no separate "enterprise" plan id. The self-hosted deployment offering described in Deploy MemDB and Self-hosted MemDB (connect) is the Teams plan; "enterprise" is a deployment style, not a different price tier.
Once a Community install passes 1,000 graph queries in a month, it doesn't stop working — it drops into a daily-allowance throttle mode instead. Any plan with no monthly limit (Teams, or a custom unlimited grant) is never gated at all.
What counts as a query#
A fixed, server-synced list of billable MCP tools is what the quota tracks — things like find_code, find_symbol, get_symbol_context, get_impact, analyze_relationships, get_evolution, get_timeline, graph-algorithm tools, quality/review tools, and API-topology tools. Community's billable list currently has exactly 37 entries.
Deliberately not billed, on every plan:
- Indexing and the file watcher —
index_directory,watch_directory, job/worktree plumbing. - Fleet coordination — every
fleet_*tool (intents, leases, episodes, escalations, audit). - Diagnostics —
embed_diag,mem_diag, and friends. - Metadata listing and stats — repo/job/worktree enumeration, directory trees, repository stats.
In short: the meter tracks insight retrieval, not the plumbing that keeps the graph fresh or lets multiple agents coordinate.
How metering works#
The billable tool list is fixed in the binary and kept in sync with the server via the same heartbeat that reports license state. Each billable call is counted locally against your plan's monthly limit; the count and the plan's limits travel together so a plan change (upgrade or downgrade) takes effect without a restart. The dashboard at localhost:3030 shows a live query-quota meter so you can see usage before you hit a wall.
Hitting the limit#
When a billable call would exceed your plan's limit, Memtrace returns a structured quota error instead of the tool result — a machine-readable shape carrying an upgrade URL (memtrace.io/pricing), not a bare failure. Your agent can surface that URL directly instead of just reporting "it broke."
Teams (and any plan configured with a null monthly limit) skips the quota check entirely — every billable call succeeds unconditionally. There is no daily throttle to fall into either.
Multi-device is fine#
Signing in with the same license on a laptop and a desktop, or across a team's machines under one seat, is explicitly supported and not treated as abuse. Billing is per query credit, not per seat or per device — a stable per-machine device id lets one account run memtrace start or memtrace mcp on as many machines as you actually use. See memtrace auth for the sign-in flow itself.
Cortex is universal#
Decision memory (Cortex) is not a paid add-on and carries no separate quota tier — it defaults to active on every plan, Community included. What Cortex retrieval calls do share with the rest of the product is the same per-retrieval metering model described above; there is no dev flag or cloud grant required to turn it on.
Next: set up self-hosted MemDB for the Teams plan, or see memtrace auth for signing in and switching plans.