context-mode

Context Mode is an MCP-focused “context virtualization” layer for AI coding agents. It provides sandbox tools to keep large raw tool outputs out of the model context, tracks editing/tool/task/error/user events in a local SQLite database, and uses FTS5/BM25-style search to retrieve only relevant past events during context compaction so the agent can resume continuity.

Evaluated Mar 29, 2026 (0d ago)
Homepage ↗ Repo ↗ Ai Ml mcp context-window session-continuity fts5 bm25 sqlite sandbox tool-sandboxing code-assistant typescript
⚙ Agent Friendliness
58
/ 100
Can an agent use this?
🔒 Security
35
/ 100
Is it safe for agents?
⚡ Reliability
26
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
78
Documentation
70
Error Messages
0
Auth Simplicity
95
Rate Limits
0

🔒 Security

TLS Enforcement
40
Auth Strength
20
Scope Granularity
20
Dep. Hygiene
55
Secret Handling
50

No explicit auth model is described (likely local execution), and no details are provided about TLS usage, secret handling, or data encryption at rest. SQLite/FTS5 indexing and sandbox execution are central features; without documented threat model, the security posture can’t be confidently rated. Dependency hygiene score is estimated from the presence of common packages but without CVE/SBOM evidence in the provided content.

⚡ Reliability

Uptime/SLA
0
Version Stability
45
Breaking Changes
20
Error Recovery
40
AF Security Reliability

Best When

You are running a tool-using coding agent for long sessions (multi-step edits, many file reads, logs, web fetches) and your current workflow suffers from context-window overflow/forgetfulness after compaction.

Avoid When

You can’t deploy a local MCP tool/plugin/hook integration or you require an external hosted managed service with explicit SLAs and compliance guarantees.

Use Cases

  • Reducing LLM context-window usage for long agent sessions
  • Improving session continuity across conversation compaction (resume with relevant edits/tasks/errors)
  • Tool-output summarization/indexing for large outputs (file reads, git operations, web fetches, shell commands, etc.)
  • Making code-editing workflows more stable by indexing intent, changes, and decisions
  • Searching previously seen artifacts/events to re-ground the agent during long runs

Not For

  • High-security environments that require strict network/data boundary guarantees without a documented threat model (details not provided in provided text)
  • Environments where local SQLite/FTS5 indexing is disallowed
  • Agents that cannot integrate MCP tools or the platform-specific hook/plugin architectures described
  • Use as a general-purpose vector search service without knowing its retrieval semantics (it’s presented as FTS5/BM25 retrieval)

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
Yes
SDK
Yes
Webhooks
No

Authentication

Methods: None described (integration appears local and invoked by the host agent/plugin)
OAuth: No Scopes: No

No authentication mechanism (API keys, OAuth, scopes) is described in the provided README content; integration appears to rely on local process invocation and agent hooks.

Pricing

Free tier: No
Requires CC: No

Pricing information is not provided in the supplied README/manifest content; npm package distribution is implied.

Agent Metadata

Pagination
none
Idempotent
False
Retry Guidance
Not documented

Known Gotchas

  • Hook/session-start availability varies by host platform (e.g., Cursor sessionStart hook rejected by validator per provided link; OpenCode/KiloCode sessionStart not yet available), which can affect resume behavior.
  • Routing/enforcement differs between platforms (MCP-only vs hook/plugin-based routing). If routing is not configured, the agent may not preferentially use sandbox tools, reducing effectiveness.
  • Local indexing implies the host environment must support SQLite/FTS5 operations; if those checks fail, the agent may degrade to less effective context handling (doctor command mentioned, but operational guarantees not shown).

Alternatives

Full Evaluation Report

Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for context-mode.

AI-powered analysis · PDF + markdown · Delivered within 30 minutes

$99

Package Brief

Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.

Delivered within 10 minutes

$3

Score Monitoring

Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.

Continuous monitoring

$3/mo

Scores are editorial opinions as of 2026-03-29.

5347
Packages Evaluated
21056
Need Evaluation
586
Need Re-evaluation
Community Powered