Perseus
Cited codebase search that grounds coding agents in exact file-and-line context before they edit.
If your agent keeps inventing file paths on a repo it cannot hold in its head, Perseus is the cheapest fix I have seen, and the $5 Demo week lets you prove it on your own code before paying anything. Multi-hop search and impact analysis are what justify the seat, since multi-file edits are where agents normally fall over. It is a retrieval layer, not an IDE, so buy it for the agent, not for yourself.
Verified 2d ago · liveness 77/100 · cite: rightaichoice.com/tools/perseus
- Developers running coding agents that need accurate, cited code context before edits
- Teams on large monorepos where agent hallucinated paths and stale state are costly
- Agent tool builders embedding a retrieval backend via the HTTP API or CLI --json
- Engineers who search code by semantic meaning — auth boundaries, routes, error paths
- Complete beginners looking for a full IDE replacement that writes the code for them
- Developers whose questions are already answered by plain file-level grep
- Shops that want a conversational repo chat product rather than a retrieval layer
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Perseus if you want a conversational repo chat product or a full IDE that writes the code for you, or if your team's agent loop would blow past 2,000 queries a month on the Developer tier.
Queries are metered per plan, so an agent loop that re-queries on every edit can burn the Developer tier's 2,000 monthly queries well before month-end.
Developer at $20/seat/month fits one developer and their agents on up to 3 repos and 2M indexed LOC; Scale at $100/seat/month is for larger codebases and heavier agent work at 15 repos and 25M LOC. Both are self-serve. The $5 one-time Demo week undercuts repo-chat and full-IDE subscriptions for a first trial, but teams with high-volume agent loops should compare the 2,000-query Developer cap against seats on Scale before standardizing.
In short
Perseus — Cited codebase search that grounds coding agents in exact file-and-line context before they edit. Best for Developers running coding agents that need accurate, cited code context before edits, Teams on large monorepos where agent hallucinated paths and stale state are costly, Agent tool builders embedding a retrieval backend via the HTTP API or CLI --json. Plans from $5.
What's new in Perseus
Checked todayAcross the latest 5 updates: 1 launch and 4 changelog entries.
Multi-hop search follows a thread across more files in one pass
Multi-hop search now follows a thread across more files in a single pass, so answers hold together across modules. Citations point to exact supporting lines with fewer stray ranges, and warm starts settle in about a second.
Incremental re-indexing of the working tree
Only changed files are uploaded, keeping the index current with uncommitted edits. The CLI also gained a --json output mode for piping ranked hits into other tools.
HTTP API endpoint for querying indexed repositories
An HTTP API returns the same ranked, cited results as the CLI, letting editors and agent integrations ground a coding agent in the code it is about to change.
Go, Rust and Ruby added to structural parsing
Go, Rust, and Ruby joined the set of languages parsed structurally rather than as plain text. Large monorepo indexing no longer stalls on generated files or vendored directories.
Private access program opened
Early users could index a repository and query it by meaning, with answers cited to file and line.
What people actually say about Perseus — is it worth it?
We scanned public community sources for Perseus on Jul 3, 2026 and could not establish that the discussion we found is about this tool rather than something else sharing its name. Our own analysis of that scan says the posts were off-subject. Rather than publish a sentiment score built on the wrong subject, we publish nothing here and re-run the scan.
Viability Score
How well maintained and how widely used is Perseus? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Cited codebase search returning exact file paths and line ranges to coding agents
- Semantic code search that finds code by meaning rather than substring matching
- Multi-hop search that follows a thread across more files in a single pass
- Constrained search that narrows a query with additional conditions
- Research packets combining 2-8 linked questions into one answer with a combined reading list
- Impact analysis mapping the dependency surface before editing a shared symbol
- Eval mode that turns the current diff into a verification plan with test commands
- Path tracing between two symbols with configurable max depth
- Diff search modes for changed, impacted, and regression queries
- Incremental re-indexing of the working tree that uploads only changed files
- perseus watch to sync edits in the background during active work
- Indexing of local working trees and GitHub repositories by URL or owner/repo
- CLI with --json output for piping ranked, cited hits into other tools
- HTTP API returning ranked cited results to third-party tools and agent integrations
- Agent skill install via npx skills add, plus perseus rules for repo-local agent rules
About Perseus
Perseus is a codebase search engine built for coding agents rather than for humans typing into a chat box. Point it at a local working tree or a GitHub repo and it returns ranked hits with exact file paths and line ranges, so an agent working in Claude Code, Cursor, or Codex reads code that actually exists instead of guessing at paths. The retrieval model mirrors how agents actually explore a repo. Semantic search finds code by meaning, multi-hop search follows a thread across several files in one pass, and constrained search narrows a query with extra conditions. For bigger questions, research packets bundle two to eight linked questions into a single answer with one combined reading list, impact analysis maps the dependency surface before you touch a shared symbol, and eval mode converts the current diff into a verification plan with test commands. Keeping the index honest is the other half. Incremental re-indexing uploads only changed files, perseus watch syncs edits in the background, and generated or vendored directories get skipped so monorepo indexing does not stall. Go, Rust, and Ruby are parsed structurally rather than as plain text. A CLI ships --json output for piping hits, an HTTP API returns the same ranked cited results to third-party tools, and the console shows the planner searching live while you watch. Pricing is tiered by indexed LOC and query volume: a one-time $5 Demo week, $20 per seat per month for Developer, and $100 per seat per month for Scale. That puts it alongside retrieval layers like Sourcegraph Cody or embeddings you wire up yourself, but Perseus is narrower by design. It does not write your code or chat about your repo; it hands an agent the cited context to do both well.
Behind the Verdict
Pick Perseus when the failure mode you keep hitting is context, not reasoning. Agents that know what to do but open the wrong file, cite a stale helper, or edit a proxy that was never the enforcement point are exactly the cases this fixes. The April 2026 HTTP API and the May 2026 CLI --json mode matter more than they sound, because they let an editor integration and a shell script read the same ranked results instead of each maintaining its own index. On cost, the math is simple. $20 per seat per month buys 2,000 queries, three repos, 2M indexed LOC, and 90 days of memory retention. Most solo developers will not approach 2,000 queries unless they run agents in tight loops, which is precisely when you should be looking at Scale at $100 for 5,000 queries, 15 repos, 25M LOC, and unbounded retention. Where it bites: this is not a conversational repo chat product, and it will not replace your editor. If your questions are already answered by plain grep, you are paying for retrieval you do not need. Teams that want a human-facing answer box will find Perseus oddly quiet. It returns citations, not prose. The closest alternative is rolling your own embeddings pipeline over the repo. That is viable, and it is also where teams quietly burn a month on chunking, incremental re-indexing, and keeping citations precise. Multi-hop search and impact analysis are the parts you would have to build yourself, and they are the parts that take longest to get right. For agent builders, the HTTP API is the interesting surface. You can keep your own planner and let Perseus own retrieval, so you are not shipping a search backend as a side quest. The tradeoff is another dependency in the loop and another place a query can fail. Warm starts settling in around a second, per the June 2026 notes,
Researching Perseus? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Perseus actually fits — and what changes day-one when you adopt it.
Index the checkout with `perseus index`, keep edits synced with `perseus watch`, then query 'auth enforcement on query route' before editing a route file.
Outcome: The agent edits the helper that actually owns the auth decision instead of adding a duplicate check in the proxy.
Index the monorepo (Go is structurally parsed and vendored directories are skipped), run `perseus impact BillingPlan --print-tests` before changing a shared symbol, then `perseus eval --run-tests` on the diff.
Outcome: You see which files depend on the symbol and run only the tests Perseus derives from the impact packet, rather than the full suite.
Index the repo server-side by owner/repo, then call the HTTP API or the CLI --json mode from your own agent harness to get ranked, cited hits.
Outcome: Your agent gets the same cited results the Perseus console shows, without adopting the console as the primary interface.
Use Cases
- Ask your coding agent to find where authentication is enforced on a query route, returning cited file and line numbers.
- Before editing a shared function, run impact analysis to see which files depend on it and what tests cover it.
- Index a large monorepo and query for specific error strings or UI labels across the entire codebase.
- Combine multiple related questions (billing route, pricing plan catalog, billing tests) into a research packet.
- Use the CLI's --json mode to pipe ranked search results into custom scripts or CI pipelines.
- Trace how a request flows from a route entry point to a persistence function with the path command.
- Run `perseus watch` during active work so the index stays current with uncommitted edits.
- Turn the current diff into a verification plan and run only the pinned tests with `perseus eval --run-tests`.
Models Under the Hood
as of 2026-10-17
Limitations
- Queries are capped by plan: 250 on the $5 Demo week, 2,000/month on Developer, 5,000/month on Scale.
- Ask calls are capped too (20 / 200 / 1,500).
- Indexed repositories are limited to 1, 3, and 15 respectively, and indexed LOC to 100k, 2M, and 25M.
- Memory retention is 7 days on Demo week, 90 days on Developer, and unbounded on Scale.
- The docs describe indexing of local working trees with uncommitted edits and GitHub URLs; no other version control system is mentioned.
- The service is cloud-dependent, with no self-hosting option in the documentation.
- The CLI docs recommend short noun phrases for queries and warn against one-word searches, and no query context window is documented.
as of 2026-10-09
Verification history
We have re-verified Perseus 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Perseus tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Demo week
$5 one-time / 7 days
Ideal for
One developer who wants to test Perseus on a single real repository before committing to a monthly seat.
What this tier adds
Starting tier: one-time $5 for 7 days with 250 queries, 1 indexed repository, 100k indexed LOC, and 7 days memory retention.
Developer
$20/seat/month
Ideal for
A developer running Claude Code, Cursor, or Codex on up to 3 repositories, including side projects plus a work repo.
What this tier adds
Adds recurring monthly access at $20/seat/month: 2,000 queries/month, 3 indexed repositories, 2M indexed LOC, 200 ask calls, and 90 days memory retention.
Scale
$100/seat/month
Ideal for
Teams on large codebases with heavier agent work that need more repos, more LOC, and longer-lived query memory.
What this tier adds
Raises the ceilings to 5,000 queries/month, 15 indexed repositories, 25M indexed LOC, 1,500 ask calls, and unbounded memory retention for $100/seat/month.
Where the pricing makes sense
The company stage and team size where Perseus's pricing actually pencils out — and where peers do it cheaper.
Developer at $20/seat/month fits one developer and their agents on up to 3 repos and 2M indexed LOC; Scale at $100/seat/month is for larger codebases and heavier agent work at 15 repos and 25M LOC. Both are self-serve. The $5 one-time Demo week undercuts repo-chat and full-IDE subscriptions for a first trial, but teams with high-volume agent loops should compare the 2,000-query Developer cap against seats on Scale before standardizing.
Setup time & first value
How long it actually takes to get something useful out of Perseus — broken out by persona, not the marketing-page minute.
Install is one curl or PowerShell line, then `perseus login` (browser on a laptop, device flow over SSH). Indexing a checkout is a single `perseus index`. First useful query is minutes after install; first value on a large monorepo depends on how long the initial index takes, and warm starts settle in about a second per the June 2026 changelog.
Switching to or from Perseus
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From plain grep/ripgrep: index the checkout once with `perseus index`, then run `perseus query` with a short noun phrase instead of a substring pattern.
- →From a conversational repo chat tool: keep asking questions in your agent, but route the lookup through `perseus query` or `perseus research` so answers come back cited to file and line.
- →From manual agent file exploration: install the Perseus skill (npx skills add) or use `perseus rules add` so agents call Perseus commands instead of walking the tree.
- →From an IDE-only search: add the Perseus HTTP API or CLI --json output to your editor/agent integration and read the same ranked, cited results.
- ↗To plain grep/ripgrep: stop `perseus watch`, drop the index, and search substrings directly — you lose semantic and cited results.
- ↗To a conversational repo chat product: export nothing; keep the repo and ask whole-repo questions in the chat tool at the cost of file-and-line citations.
- ↗To a full IDE agent: point the IDE's agent at the repo directly and use its built-in context tools instead of the Perseus API or CLI.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Perseus”, and we withheld 6: 6 could not be judged, because “Perseus” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Perseus.
Official links
Tools that pair well with Perseus
Common stack mates teams adopt alongside Perseus, with the specific reason each pairing earns its keep.
Magic.dev
Magic builds frontier code models with ultra-long context to automate software engineering and AI research.
Claude
Claude is Anthropic's AI assistant for long-document analysis, agentic coding, and multi-step work inside one chat and Cowork surface.
Imbue
Imbue builds inspectable coding-agent tools, from mngr's parallel agent runner to Latchkey's HTTP permissions
Featured Head-to-Head Comparisons
Perseus vs Spider Cloud
Spider Cloud and Perseus solve completely different problems: one fetches live web data for AI agents, the other indexes code for coding agents. Choose Spider Cloud if you need fast, cheap web scraping with AI extraction and browser automation. Choose Perseus if you need accurate, cited code search for your agentic workflows. No direct competition.
Perseus vs Voyage Ai
If you're building enterprise RAG pipelines requiring high-accuracy retrieval on domain-specific text—especially finance, legal, or code—Voyage AI offers superior embedding and reranking models with long-context and low-dimensional options. However, if your focus is grounding AI coding agents in real code with cited, multi-hop search, Perseus is the specialized tool, especially with its recent speed and citation improvements. Choose based on your primary need: document retrieval versus code-native search.
Perseus vs Temporal Ai
Temporal AI and Perseus serve completely different needs: Temporal is for building reliable, long-running AI agents and workflows with automatic failure recovery, while Perseus is a code search engine that grounds coding agents in accurate context. If you need to orchestrate multi-step agentic workflows that survive crashes, choose Temporal. If you want your coding agent to retrieve precise, cited code snippets quickly, go with Perseus. They are complementary – you could even use both together.
Alternatives to Perseus
View all
Frequently Asked Questions
Best-of guides
Used Perseus? Help shape our editorial sentiment research.